Version 1.7.4
Version 1.7.3
Version 1.7.2
Version 1.7.1
Version 1.7.0
Version 1.6.5
Version 1.6.4
Version 1.6.3
Version 1.6.2
Version 1.6.1
Version 1.6.0
Version 1.5.1
Version 1.5.0
Version 1.4.1
Version 1.4.0
Version 1.3.3
Version 1.3.2
Version 1.3.1
Version 1.3.0
Version 1.2.9
Version 1.2.7
Version 1.2.6
Version 1.2.5
Version 1.2.4
Version 1.2.3
Version 1.2.2
Version 1.2.1
Version 1.2.0
Version 1.1.6
Version 1.1.5
Version 1.1.4
Version 1.1.3
Version 1.1.2
Version 1.1.1
Version 1.1.0
Version 1.0.9
Version 1.0.8
Version 1.0.7
Version 1.0.6
Version 1.0.5
Version 1.0.4
Version 1.0.3
Version 1.0.2
Version 1.0.1
|
version 1.7.5, in progress
- max() and min() functions: extend the syntax to allow a list
argument
- Reading CSV data: automatically substitute '_' for illegal
characters in variable names
- Support DESTDIR for make install
- Add Weibull distribution to the various gretl probability
distribution functions
- Fix for parsing a restriction with leading numerical term
- New functions: pdf, sdc, msortby
- User-defined functions: allow a descriptive string to
accompany each parameter, for use in GUI
- Add $unit accessor to get a panel unit index
- Add "--breusch-pagan" option to lmtest command (with --robust
option to use the Koenker robust variance estimator); also
add this test to the GUI
- Fix some bugs in sub-sampling of panel datasets
- Revise the GUI for setting the sample range for panel data
- Improvements for panel time-series graphs
- Fix bug in importing data with long numeric observation
labels
- printf command: allow matrices in place of scalars (e.g.
using "%f" or "%g" conversion)
- GUI script output window: add option of making output
"sticky" (running a script appends to text)
- Add log-axis option to GUI graph editing dialog
- GUI program: add right-click popup menus to more windows
- Fix bugs 1929957, 1922915, 1951681, 1954428, 1954838
- Add guard against unmatched "if" in loops, functions and
scripts
- Fix bug: VARs and equation systems not getting removed as
"last model" on exit from user-defined functions, which
could cause strange errors in complex scripts
- Fix for forecasts from equation systems displayed in the
GUI program
- Updates to some screenshots in the User's Guide
- pca command: add --covariance option to base the analysis on
the covariance matrix instead of the correlation matrix
- corr command: document the --uniform option and add it to
the GUI
- Fix bug: loops inside conditionals inside a loop were not
executing properly in some cases
- X-Y scatterplot: add option of controlling for a third
variable
- hsk command: fix for the case where the model includes a
dummy that is specific to a particular observation
- arima specification dialog: add checkbox for basing
standard errors on the Hessian
- store, outfile, etc.: if the filename starts with "~/",
expand this to the user's home directory, if possible
- Search for data file collections: add the user's gretl
data directory ("dotdir") to the path searched
- "append" command: try harder to make this do what the user
is most likely to want in the case of panel data
- frequency distribution in GUI: allow choice of bin size,
as with frequency plot
- fixed effects estimation: print the average constant
- eigengen function: make this behave as the manual says
- OLS anova: add p-value for the F-statistic
- Fix VECM forecast graph in GUI ("pre-forecast" fitted
values were not integrated)
- "open" command: add options for controlling importation
from spreadsheets
- add "set debug" option for deugging user-defined functions
- GUI function package: add option to save the packaged
functions as a regular gretl script
- Fix bug in modification of LMF autocorrelation test (bug
was introduced in gretl 1.7.4)
- Script editor: add option to show line numbers
- Native database windows: add right-click popup menu
- Local function package listing: sort alphabetically
3/21/2008 Version 1.7.4 [Back to top]
- Fix for ADF-GLS test with missing values
- Fix for detection of unit-diagonal matrix (oops!)
- Merge fit and fcasterr commands with fcast
- Add $fcast and $fcerr accessors for use with fcast
- Autocorrelation test: use Kiviet's original 1986 formula
for the LMF statistic
- "system" command: enable naming of systems in the manner
"sysname <- system"; enable forecasts; make parsing of
identities more robust; enable TeX view/print in GUI;
add new accessors $yhat, $sysGamma, $sysA, $sysB
- "estimate" command: add --quiet option
- remove obsolete command: multiply
- Consolidate the commands corc, hilu and pwe into one new
command with options: ar1
- Fix for the case of graphing 6 time series together
- Fix bugs 1908139, 1909469
- Fix documentation of $stopwatch
- Move handling of strings into "genr"; add strlen function
- fdjac and BFGSmax functions: don't require that the
function-call argument be wrapped in quotes
- "corr" command: allow printing of ranked data for Kendall's
tau as well as spearman (verbose option); also add a
verbose check-box to the rank-correlation dialog
- User's Guide: add a discussion of Arellano-Bond
- Enable "delete" command for saved strings
- Enable naming of matrix columns, via colnames function or
the GUI matrix editor
- Enable "dataset resample" in loops (experimental)
- Runs test: enable the "--difference" option in the CLI;
handle the case where positive and negative values are not
equiprobable by default, but add an "--equal" option to
impose the equiprobability assumption
- Fix for "data error" in VECM autocorrelation tests (for now
we'll just do Ljung-Box)
- Fix sub-sampling bug (new in 1.7.3)
2/29/2008 Version 1.7.3 [Back to top]
- Fix for 64-bit platforms: don't use "long" as the C
counterpart of Fortran 77 "INTEGER"
- Make the rcond function available for asymmetric matrices
- Enable the $yhat accessor for fitted values in equation
systems
- Add Italian function help to Windows build
- Weekly to monthly data compaction: respect the user's
choice of compaction method, as claimed
- Add trimr matrix function (as in Gauss)
- Extend matrix "division" a la Gauss
- Disallow user functions with the same name as a built-in
function
- Update several screenshots in User's Guide
2/25/2008 Version 1.7.2 [Back to top]
- list-creation: add "dataset" option to use all currently
available data series
- Fix bug in recording result of TSLS heteroskedasticity test
- Add sscanf command (a subset of C sscanf functionality)
- Add shell capture for strings via $(shellcommand)
- Paths stuff: add (non-configurable) "dotdir" to hold per-user
automatically-created files
- Fix bugs 1861010, 1868859, 1869271, 1870459, 1870973,
1871110, 1871135, 1873625, 1874601, 1881849, 1897648
- Rationalize frequency plot dialog
- Add Help item to forecast dialog
- Fixes for matrix "dot operations"
- Fix breakage in graph color selector
- Fixes for ODS spreadsheet importer
- Document the --auto option for the "omit" command, and add
a GUI handle for it
- MS Windows and OS X packages: update to gnuplot 4.3 (CVS)
- Graphing internals: use UTF-8 for internationalized strings
- Models having an exact fit: block various invalid tests
- det() function: fix for singular matrices
- GUI script editor enhancements
- Allow drag of database from databases-on-server window
to local native databases window; also function packages
from server to local function package window
- Add basic support for the DF-GLS test via a new --gls
option to the adf command
- Add selifc and selifr matrix functions
- Generalize several functions to accept matrix as well as
series argument, e.g. quantile, diff, cum, resample
- Add new accessors for TSLS test results: $hausman and
$sargan
- Many small improvements to the GUI
- Add systematic documentation for the functions available
in the genr command
- Enable "append" command in loops
- Ordered logit and probit: don't insist that the dependent
variable has a minimum of zero
- Windows installer: update GTK version, packaged zlib1.dll,
and gtksourceview DLL.
- Fix bug in arima auto-differencing, for non-seasonal order
greater than 1 plus seasonal differencing
12/25/2007 Version 1.7.1 [Back to top]
- Add documentation for "set stopwatch"
- Add movavg function to genr (not documented yet)
- Add autocorrelation and heteroskedasticity tests for models
estimated via TSLS
- Add "dataset resample n" command to resample observations
with replacement (not documented yet)
- Add functions for matrix I/O from/to text files
- Fix garch bug when p!=q
- Fix bugs: 1848487, 1847908, 1850842, 1848039, 1840179,
1842439, 1841098, 1843664, 1835707, 1841107, 1841774
- Fix bugs in saving/re-opening restricted VECM models as part
of a gretl session
- Fix bug in CUSUM/CUSUMSQ plots for time series data frequency
greater than 1
- Work around strange behavior of gnuplot 4.2 on Ubuntu
- Reinstate user-level choice of "Windows-look" emulation
- Reinstate user choice of menu font on Windows
- Script editor: add right-click popup with menu items to
comment/uncomment selected block of commands, and to execute
a line or region
- genr: fix for case where closing parenthesis is missing at
the end of a formula
- OS X: replace fftw library with an i386 build
- Toolbar: use GTK stock icons where possible
- Gnuplot commands editor: fix behavior of save button
- Time series plot, more than one series: add option to show
separate small plots
- Output from script execution: reuse existing output window
if it's left open
- Add set variable "fcp" to use Fiorentini, Calzolari and
Panattoni GARCH code
- Add set variable for max iterations in a "while" loop
- Fix bug: missing data handled incorrectly in ascii graphs
- Removed obsolete genr function varnum()
- Renaming session objects: move to popup menu
- Saving text from output windows: give choice of saving to
file, or as session icon
- Commands allowable in loops: enable all but a fairly small
subset
- Shell command: when invoked in gui console, send output to
gui console
- Fix handling of "quit" when it's blocked by "if"
11/27/2007 Version 1.7.0 [Back to top]
- Add heckit estimator
- Add estimation/testing for restricted VECM models
- Fix for loop on empty list
- Fix for remaining issues with apostrophes in graph titles
- Fix for lists as arguments for user-defined functions
- Fix for saved sessions after using Forecast menu items
- Fix for long variable names in Model Graph menus
- Improve formatting for nonparametric difference test
- Improve formatting for "freq" command; also add --silent
option for this command
- Add ceil() and floor() functions in genr
- Add "--p-values" option to show p-values instead of slopes
for logit, probit models
- Add options to GUI for xtab command (cross-tabulation);
also add Fisher's Exact Test for 2 x 2 tables
- Fix for graphing the time trend, and for graphs displaying
more than 8 series
- Fix for length of correlogram on weekly data
- Fix for null matrix arguments to user functions in GUI
- Fix for floating point values in "set" when using
decimal comma
- Fix for opening cross-sectional gretl data sets in R version
2.5.0 or higher
- Fixes for autoregressive genr formula: speed-up, and fix bug
when such formulae are used in an nls/mle block
- Double maximum length to gretl command line to 8192 bytes
- Gamma p-values: switch to shape/scale parameterization
- Spearman's rho: fix for case with ties
- Allow for syntax change re. missing values in gnuplot 4.3
- "corr" command: add --spearman and --kendall options
- Rank correlation in GUI: add Kendall's tau option
- Fix a few issues with extreme values in cdf, pvalue code
- add gamma() function in genr to generate gamma random
variable
- Modify Locke's test to use Kendall's tau
- Add ginv() (Moore-Penrose inverse) matrix function
- Add multivariate Portmanteau test for autocorrelation in
VARs (Ljung-Box)
- Allow concatenation of multiple short-form options, as
in "-rdq"
- Modify retrieval of info from VARs: $coeff gets the matrix
of coefficients, $compan gets the companion matrix
- New accessor: $gmmcrit gets the GMM criterion for a model
estimated via GMM
- gretlcli: Ctrl-X exits without prompting for save
- Databases: allow selection of more than one series for
display, graphing, importing
- Databases: allowing deletion of series from native gretl
databases (if the user has write access)
- New command, "dataset": subsumes the old "addobs" and
"transpos", while adding new options
- fftw DLL for Windows: fix a problem experienced on some
Windows systems
- Generation of random variables in GUI: use an interface
consistent with that for p-values, critical values, etc.
- Fix for XLS importer with non-zero column offset
- Improve error detection and recovery for adding case
markers from file
- Graphs: try to respect line color choices when exporting
to postscript and PDF
- Small improvements to built-in data editor
- ARMA: allow specification of particular AR and/or MA
lags (int the non-seasonal component only)
- Add importer for Open Document Spreadsheet files
- MS Windows: update to version 2.12 of GTK runtime
- MS Windows: update to gnuplot 4.2.2, and add switch
in graph editing dialog to prevent anti-aliasing of
lines
- GUI linear restrictions dialog: fix bug whereby
parameter names were not accepted
- Saving databases: fix effect of --overwrite flag to
do what the manual says
- Databases on Windows: look in the gretl user directory
as well as the system db directory
- add argname() function for use in user functions
- Panel, unit-weights: fix omit command
- Fix bug in mailer dialog (crash when closed via
window manager)
5/17/2007 Version 1.6.5 [Back to top]
- Fix new breakage in configuration of user directory
- Updates to manual
- Mark some more strings for translation
- Add new menu set /Tools/Nonparametric tests, with GUI
access to the "difftest" and "runs" commands
- Add --difference option for runs command
- Various small bug-fixes
5/13/2007 Version 1.6.4 [Back to top]
- Fix breakage in relation to the Windows registry
- Fixes for accented characters in graphs, on systems with
UTF-8 as the standard character encoding
- Additional fixes for Polish text in graphs
5/10/2007 Version 1.6.3 [Back to top]
- Add facility for bootstrap analysis in GUI, for OLS
- Model table: fixes for including panel models; allow up to
12 models; formatting enhancements
- Fix bug in panel autocorrelation test (lmtest command)
- Add fft() matrix function (Fast Fourier Transform)
- Add cmult() matrix function (complex multiplication)
- Add quantile() function in genr
- Add production of restricted estimates to the "restrict"
command as applied to VECMs
- Multiple fixes for session icon handling
- LAD models: print the median of the dependent variable
- Matrix properties window in GUI: add an indication of whether
or not a square matrix is idempotent
- Panel data, pooled OLS model: when robust standard errors are
chosen, use the Arellano-type HAC estimator (well, it's HAC
when the time-series dimension is small); also offer Beck-Katz
Panel Corrected Standard Errors
- Apply the above to tsls in a panel context also
- Panel data, "summary" command for single variable: show the
within and between standard deviations
- Add "--quiet" option to lmtest command (don't print the
auxiliary regression)
- Observation labels: ensure these are properly XML-encoded when
saving a data set as gdt
- Fix bug that could cause a crash when exiting the program on
MS Windows
- Fix bug: crash when saving a graph as an icon, when the icon
window is currently displayed
- Fix an encoding issue with the command-line help files on
platforms where the default encoding is UTF-8
- Correlation matrix: try to be smarter in face of non-aligned
missing observations
- CSV data: add support for Mac-style text files
- "loop" command, "progressive" option: fixes for the case
where this is invoked for loops other than the simple
"count" type
- Further optimization of gretl_matrix functions
- Various modifications to the configuration process
3/24/2007 Version 1.6.2 [Back to top]
- Add Poisson distribution to p-value finder
- Fix breakage in GUI "graph page" apparatus
- Graph page: give color / monochrome option
- Bivariate scatterplots: give the option of showing various sorts
of fitted line: linear, quadratic, inverse or loess
- Monochrome EMF graphs: make sure point types are distinct for
different series
- Internal clean-up of the graphing code
- Make frequency plots more configurable via the GUI
- Fix memory leaks in GUI graph editor dialog
- Bugfix: estimating models inside nested functions could cause
a crash
- Session icon window: don't display temporary matrices,
generated within functions
- Icon window: fix display on one-character object names
- Fix memory leak when generating some sorts of matrices
- Add option of using Parzen or Quadratic Spectral kernel (the
default is Bartlett) when computing HAC covariance matrix
- Add a toolbar to windows that display lists of files
- Several fixes and enhancements for function package editor
- Fix for "omit" command in relation to arbond estimation
- Correlation matrix, etc, in GUI: don't disable menu items
if only one variable is selected
- System estimation: fix breakage of the "save=" mechanism
for residuals and/or fitted values
- gmm: improve syntax handling for orthogonality conditions
- Update the API documentation (still very much incomplete)
- Update gnuplot to version 4.2.0 in the Windows installer
2/26/2007 Version 1.6.1 [Back to top]
- Revamp for internal "genr" mechanism plus some user-visible
changes to genr
- Make the "delete" command do what the manual says
- Doing "Edit attributes" on a newly generated variable: you
can edit the formula and hence change the content of the variable
- Displaying table of summary statistics for multiple variables:
allow slightly longer variable names before falling back from
combined tabular format to individual tables
- Sorting variables in main window: always revert to sort by ID
number after making changes (e.g. loading a new dataset)
- Main-window variable listing: don't always jump to line 1 after
making changes
- When you "Save as" in the Command log window, ensure that all the
regular script-editing signals are properly connected (e.g. the
state of the Save icon responds appropriately)
- "dummify" command: add options --drop-first and --drop-last; make
--drop-first the default when the dummify() construct is used in
a regression list
- Add ordered probit analysis for the case where the dependent
variable is discrete but not binary
- Add computation of generalized residuals in various estimators
- Revamp Tobit estimation: use BFGS instead of BHHH, allow --robust
option and add normality test
- Fix some bugs and inconsistencies in relation to the Ljung-Box
statistic and ACFs
- Fix bug in the Baxter-King filter routine
- Add the CUSUMSQ test alongside CUSUM
- Add test version of Arellano-Bond dynamic panel data estimation
- Fix bug in GUI spreadsheet in relation to adding variables
- Fix bug in restricting the sample in the GUI, for the case
where's there's already a restriction in place
- QLR test: add time-series graph of F test
- new command "xcorrgm" for cross-correlogram of 2 variables
- Add bivariate normal to cdf()
- New modular engine for garch estimation (no user-visible changes
for now)
- Make window size for semiparametric estimators of the fractional
integration parameter adjustable by the user
- Make the number of bins for a frequency plot adjustable via GUI
- Add facility to generate Binomial and Poisson random variables
- Row format for tabular output of models in TeX: make this
configurable
- Add several matrix functions
- Add nonlinear GMM estimation (experimental)
- Numerous small bug-fixes
9/13/2006 Version 1.6.0 [Back to top]
- Revamp mechanism for saving sessions in gui program (we
now save sessions as zipfiles containing XML representations
of any models that have been estimated)
- Add "native" exact ML estimation of ARMA models using the
Kalman filter
- Saving data in native gretl format: make gzip compression the
default
- Add some basic functionality for analysis of discrete variables.
New behavior of frequency distribution, plus new commands:
"discrete" (interpret variables as discrete), "dummify" (create a
set of dummies coding for the values of a discrete variable), and
"xtab" (cross-tabulation of two discrete variables)
- Don't add an extra newline at the end of plain text data files
(e.g. CSV)
- "restrict" command for single equations: add a printout of the
restricted coefficient estimates and standard errors
- Add mechanism for packaging user-defined functions so as to
make these available for other users; also add a mechanism
for downloading such packages from a server, and for uploading
new packages to the server
- Redefine the weight variable in gretl's "wls" command, for
better compatibility with other programs: the weight var is
now defined as the series of diagonal elements of the weighting
matrix, so existing gretl scripts that use wls should square
the old weight variable
- Don't print the "model has no constant" message, and do not
use special methods to calculate R^2 and F, if the set of
regressors spans the space of a constant
- Indexing of coefficients in "restrict" command: standardize on
using a 1-based index in all contexts
- Revamp panel data estimation: redefine the "panel" command for
fixed- or random-effects estimation, rename the old "panel"
command as "paneldat"; refine calculation of Hausman test statistic
- Lags in main window: make these visible under their parent
variable, via a "twisty" triangle
- Revamp multiple-precision models: add more statistics, and make
them available for capture via genr
- Revamp test for differences of two means in the GUI hypothesis
test calculator
- Imposing panel structure on a data set: add the option of
defining the structure by reference to two index variables, one
for the unit and one for the period -- this is implemented in
the "setobs" command and the GUI data structure wizard
- Add --robust option to nls command (nonlinear least squares)
- Add --hessian and --robust options to the mle command, which
respectively produce a covariance matrix based on the Hessian
and a QML sandwich estimator of the covariance matrix
- Add Quandt likelihood ratio (QLR) test for structural break
- ADF and KPSS tests: add option to difference the selected
variable before testing
- Delimiter for "CSV" files: add semicolon to the list of
options
- VECMs: make Johansen's alpha and beta retrievable as matrices
with identifiers $jalpha and $jbeta
- Several bug-fixes for the matrix manipulation facility, plus
a few new functions (meanr, meanc)
- GUI menus restructured and polished
- Add facility for multiple time-series graphs from the GUI
- Revamp the "--ten" option for printing variables to 10 digits:
replace with a configurable "--long" option
- Graph editing dialog: make the line width configurable, and
also make the preferred line width settable per variable in
the "Edit attributes" dialog
- Data import and export: add option for reading and saving data in
JMulTi format
- Databases: add capacity to read PcGive in7/bn7 databases
- Gretl command minibuffer: model commands now produce GUI
model windows (for single-equation methods)
- QR decomposition option: fix the F-statistic values produced
for autoregressive models
- pvalue() function: always return the right-hand value, but add
a complementary cdf() function that always gives the left-hand
value
- NIST linear regression test suite: re-express accuracy in terms
of log relative error, plus other enhancements
- Generating random variables: add option to generate t and
chi-square values in GUI; add option of setting the PRNG seed
to the GUI variable-generation dialogs
- Time-series filters: Add Hodrick-Prescott and Baxter-King filters
as a GUI menu item, along with simple moving average and
exponential moving average
- Engle-Granger cointegration test: improve printout, add option
to test down for the lag order, and fix an issue with the
selection of the MacKinnon p-value for this test
- Tobit: enable handling of missing observations for this estimator
3/24/2006 Version 1.5.1 [Back to top]
- Add matrix manipulation functionality
- Fix translation bug for VAR menus
- Fix bug in assignment of values returned from user-defined
functions
- Small fix for helpfile formatting
- Add calculation of Gini coefficient and Lorenz curve to GUI
Variable menu
- Provide option of including seasonal dummy variables in
Dickey-Fuller tests
- Add two more "genr" functions: qnorm (quantiles of normal
distribution) and gini (Gini coefficient)
- Fix some bugs in handling of named lists, and start documenting
this properly
- Enforce use of '.' as decimal separator in genr formulae
- Fix filename encoding issue with opening of CSV files
- Eliminate deprecated command "sim"
- Extend maximum length of variable names to 15 characters
- Disable "shell" command by default (it can be enabled under
/File/Preferences)
- Change the names of the internal variables coeff, stderr and
rho to $coeff, $stderr and $rho
- Add various keyboard shortcuts to main window, including
Alt-X for a "gretl minibuffer" a la Emacs
- Time series model menu: add option for selecting optimal
lag length for a VAR, based on information criteria
- VARs: add a "Tests" menu item to GUI VAR window, permitting
tests for omission of exogenous variables (if any)
- Model window, Graphs menu: if the data are time series, add
options to view the correlogram and periodogram of the
residual from the model
- Add Hannan-Quinn criterion in all contexts where Akaike and
and Schwartz criteria are shown
- Various fixes for model selection dialog
- win32: fix the issue whereby copy-and-pasted EMFs had a
large margin to the top and right
- win32: fix opening of files with non-ascii characters in their
names
- XLS and Gnumeric importers: supply automatic varnames if no
variable names are present
- ARMA: Alter presentation of parameter estimates (call them
"phi" and "theta"); provide native multiplicative seasonal
models
- Fix issue with spectrum plot when the order of magnitude
of the series is very low
- Introduce "set" parameters for tweaking the BHHH algorithm
convergence criteria
- Add GUI lag selector
- VARs: add numerical output to VAR roots plot
- Allow commands like var and arma to be specified with a scalar
variable as order
- Add "dsort" command for reverse sorting
- Enhance and robustify script capabilities
12/15/2005 Version 1.5.0 [Back to top]
- Add VECM (Vector Error Correction Models) for time-series
data
- Add confidence ellipse option for two coefficients, in
model window menu
- New command "mle" : a general-purpose maximizer of log
likelihood, using the BFGS algorithm -- works much like the
"nls" command, taking a user-specified log likelihood function
as input, along with (optionally) derivatives of the log
likelihood with respect to the parameters
- More XLS importer fixes
- Add support for seasonal ARMA; add "sdiff" command (and
corresponding function in genr) for creating seasonal
differences
- Add capacity to produce time-series graphs with panel
data sets
- Accommodate custom time-series frequencies
- Johansen cointegration test: extend to cover all 5 cases,
produce a fuller report (loadings, long-run matrix)
- Add bootstrapped confidence intervals for impulse responses
in VAR models
- Add test for multivariate normality of VAR residuals
- Add autocorrelation and ARCH tests for VAR equations
- Add printing of cross-equation covariance matrix for VARs
- Add combined plot of residuals for VARs
- Add combined impulse response plot for VARs
- Add log-likelihood, AIC and BIC for VARs
- Add inverse roots plot for VARs (Jack Lucchetti)
- Add log-likelihood for models estimated via OLS
- Add "robust" option for logit and probit (compute QML
or Huber-White covariance matrix)
- Add "robust" standard errors option for WLS
- Remove overly verbose notification dialogs
- Add importers for Eviews workfiles and Stata .dta files
- Editable windows: show Save button as disabled if there's
nothing to be saved
- Model selection dialog: ensure that the selected regressand
is not also listed among the regressors
- Graph editing: make several more types of graph editable
- Graph saving: add PDF option if it's supported by gnuplot
- Main window: don't show generated lags in the list of variables
- ARMA: remove the ARMA item from the Variable menu (it's still
there in the Model menu); revamp the X-12-ARIMA driver so it
can handle models with exogenous regressors
- gretl graph page: add option to save as TeX source
- TeX: enable pdflatex and viewing of TeX output in PDF form
(to use pdflatex, enter this as the name of the program to
compile TeX files under /File/Preferences/General/Programs;
this is now the default on Windows)
- Add "Send To" (email) functionality for data files and scripts
- "store" command: take file type from filename suffix for
".R" (GNU R format), ".m" (Octave format) and ".csv" (comma-
separated)
- Add new "set" variable, "csv_delim", which sets the delimiter
used when exporting data as CSV -- valid options are "comma"
(the default), "space" and "tab"
- Add a new option for CSV export, "--omit-obs": suppresses
the printing of a first column containing the observations
- X-Y graphs: display equation of fitted line when a fitted line
is shown
- X-Y graphs: when fitted line is shown, add "OLS estimates"
menu item to see details of the simple regression
- Graphs with labeled data-points: give option of showing all
labels (for n < 55)
- Windows displaying two to five data series: add toolbar option
to sort the data, with choice of variable by which to sort
- Test statistic calculator (Utilities menu): besides the
entry of numerical values for sample statistics, enable the
calculation of sample statistics based on selection of variables
from the current data set, with or without sub-sampling
restrictions
- Revamp the "online" (plain text) help system
- Split gretl manual into a User's Guide and a Command Reference
and update both parts
6/30/2005 Version 1.4.1 [Back to top]
- Fix horrid unnoticed breakage in bhhh_max (broke both arma
and tobit, big mess)
- Add utf8 validation/conversion for database descriptions
(Banco de Espana databases)
- Allow dynamic forecasts with models estimated via corc and
similar
6/29/2005 Version 1.4.0 [Back to top]
- Revise definition of functions in gretl scripts (a backward-
incompatible change) in the interest of more structured
programming
- Introduce named lists as a type that can be created and
manipulated
- Big revamp of forecasting code. Make forecasts available for
all types of models; give a choice of static versus dynamic
forecasting where relevant; make forecast standard errors
available in more cases
- Increase max length of gretl script command line to 4095 bytes
- Add support for decennial data
- Fix for finding TrueType graph fonts on Windows 2000
- Add new "set" option: "horizon", to configure horizon for impulse
responses in context of VARs; also add dialog box to set horizon
for impulse response graphs in GUI
- Disable autocorrelation and ARCH tests in gui model test menu,
in case of cross-sectional data
- Add Local Whittle Estimator for fractional integration, in context
of periodogram command
- Add fracdiff function (fractional differencing) to the genr
command
- sort function in genr: save sorted case markers if the dataset
has case markers
- Bug fixes for data editing spreadsheet
- Another bug fix for the XLS importer
- In correlogram (acf) output, show calculated 5% critical value
- Correlogram: acf calculations should not use more than 20 percent
of the available data by default (Tadeusz Kufel)
- Fix bug in Tobit command in the gretl GUI
- "omit" command: add asymptotic Wald test for joint significance
of omitted variables, for estimators other than OLS
- ADF test: allow greater lag length if the data permit
- Add facility to import GNU Octave ascii data files
- Improvements to formatting in Spanish translation
- In statistical tables utility, add a bit more information for
Durbin-Watson and improve formatting
- smpl --restrict: make it not an error if the restriction has
no effect (drops no observations)
- Confidence interval for coefficients, GUI model menu: make
available for all models
- Add matrix of actual and predicted outcomes to the printout
of results for discrete models
- Improvements to many dialog boxes
- Numerous small bug-fixes
3/14/2005 Version 1.3.3 [Back to top]
- Add "poisson" command (thanks to Jack Lucchetti)
- Fix memory leak in VAR command with impulse responses, in the
command-line program
- Fix bug: possible crash in "loop foreach" if an invalid variable
name is given
- Fix bug: "display data" broken in gui program for long datasets
("press a key to continue" mechanism was invoked erroneously)
- Fix bug: wrong variable shown when you select "Display actual,
fitted, residual" for a GARCH model
- Fix bug: hausman test for pooled model would crash if there
were not sufficient degrees of freedom to estimate the group
means model
- Fix bug: crash if default data and script files not installed
with program (thanks to Ignacio and Dirk for finding this)
- Fix bug in genr: unary minus could be wrongly handled within the
argument to a function
- Fixes for appending data (make more flexible: allow appending of
observations and/or variables)
- Fixes for cointegration tests: improve dialogs and prevent crash
on selection of just one variable plus the constant
- Fix naming of Mahalanobis distances variable
- Respect omission of constant from an ARMAX model
- Graph control dialog: add one-click option to force use of a
single y axis
- New: in case of AR(1) estimation (Cochrane-Orcutt and similar), if
a lagged dependent variable is included then we automatically
use the procedure set out in Ramanathan, 5e, p. 450, to generate
consistent estimates of the standard errors
- Fix: gui menu items for test of equality of means, with equal
or unequal variances: these were hooked up in reverse
- corc, hilu and pwe AR(1) commands: make these available in
loop mode
- add "quiet" option for the commands adf, kpss and wls
- add "simple-print" option for ols command (just print the
coefficient table, with no auxiliary statistics)
- add "mahal" command: calculate Mahalanobis distances given
a set of variables; also available in GUI via right-click
popup menu when more than one variable is selected in the
main window
- genr command: $nvars gives the number of variables in the
dataset
1/28/2005 Version 1.3.2 [Back to top]
- Fix bug in auto-determination of lag in ADF test: was looking
at the wrong coefficient in some cases
- Fix bug in VAR impulse responses: could get messed up by
missing values that differ across series
- Fix bug: crash on generating LaTeX tabular output for a model
estimated using robust standard errors (format string issue)
- Fix bug: the session "graph page" was not appearing even when
a valid latex executable was available
- Fix bug: gretl console would not respond in case of an error when
executing a command loop
- Fix bug: the "restrict" command failed in the GUI program when
not applied to a named equation system
- Fix bug: printing "-inf" for AIC and BIC in case of a perfect
linear fit
- Fix various issues in relaton to nested loops (for example, where
the control variable for a given loop is not yet defined at compile
time, but will be defined in an embedding loop)
- Fix sort-of bug: possible runaway generation of graphs in response
to some loop commands
- Improve efficiency of test for existence of helper programs on
Linux/unix
- Add auto-lag option to the GUI dialog box for the ADF test
- Remember the options selected in the ADF dialog from one invocation
to the next
- Help files: include menu-path information in the GUI help file as
well as the script commands help
- Provide a mechanism for retrieving the value and/or p-value from
the last statistical test performed (via the special variables
"$test" and "$pvalue" in genr)
- Perfect collinearity: try to detect this, and remove one or more
redundant variables from the regression list if need be
- "label" command: if only a varname is given, print the variable's
current label, if any
- "genr unit": generates an index variable for the cross-sectional
units in a panel dataset (panel data only)
- Add "hurst" command (and GUI menu item) to compute the Hurst exponent
(a measure of long-term persistence used by Mandelbrot) for a given
series
- Remove unused GUI popup menu item "Simulate..."
- Updates to documentation (manual and helpfiles)
1/19/2005 Version 1.3.1 [Back to top]
- Add new dataset structure "wizard" in place of the old dialogs
for setting frequency, starting observation and time-series
or panel interpretation of the data
- Add FIML and LIML estimation to system command; also add
LR over-identification tests for these systems
- Add Hansen-Sargan over-identification test for SUR and 3SLS
system estimates
- Add "--iterate" option for estimation of systems via SUR and
3SLS
- Add "estimate" command for equation systems, and modify the
"system" command to allow the user to define named systems
- Add gui configuration for robust standard errors options
- VAR dialog: add options for robust standard errors, and for
the printing of impulse responses
- Add univariate kernel density estimation (under the Variable
menu in the gui program -- no script counterpart)
- Purge the lesser-used model selection statistics, and use
the original formulae for AIC and Schwarz's BIC
- Improve feel of data editing spreadsheet, gtk2 version
- Model dialog box: allow shifting of regressors with middle-
click of mouse
- Add groupwise heteroskedasticity test for panel data
- Extend handling of missing observations in panel context,
including case of unbalanced panels (to some degree)
- Fix for appending observations (built-in spreadsheet): plotting
vars were not being automatically extended
- Fix for deleting variables: don't show the "variables have been
renumbered" message if the only vars to be renumbered are
hidden ones
- Fix newly introduced bug in time-series graphs: don't show two y
axes if only one is actually used
- Fix for johansen cointegration test: don't crash if the user
puts the const into the list of variables
- Fix bug in forecasts with standard errors, for case of missing
observations
- Fix bug in reading of CSV files: crash if variable names were not
found, but observation markers were found
- Add new option under Sample menu, "Transpose data" (for data that
were read in "sideways" from, e.g., a CSV file
- Fix bug in ADF p-values for locales which do not use '.' as
the decimal point
- Add BIC (Bayesian Information Criterion) to the list of model
statistics that can be retrieved via genr (internal variable $bic)
- Add Baxter-King bandpass filter (bkfilt function in the "genr"
command), thanks to Jack Lucchetti
- Add Kmenta (1986) dataset and script, for a further example of
system estimation
- Add script to replicate Perron (1997) testing for unit roots
with a structural break at an unknown point
- Add an emacs file, utils/emacs/gretl.el, offering an emacs
major mode for editing and running gretl scripts.
11/15/2004 Version 1.3.0 [Back to top]
- add "append" command, to append data to the current dataset,
as in the gui File/Append data
- add KPSS stationarity test ("kpss" command, Jack Lucchetti)
- fix bug: data importers were accepting duplicate names for
variables
- XLS importer: handle string formulas; fix handling of large
string tables (SST and CONTINUE records)
- fix bug: error in printing of model selection statistics for
very large or small values on Windows
- fix bug in running scripts from the gretl console: file was
not closed when an error was encountered
- fix bug in Prais-Winsten estimator
- fix bug in influential observations printout, for non-
English locales
- revamp the Dickey-Fuller test functionality, with more
options, clearer presentation (I hope), and p-values
using James MacKinnon's code (JAE, 1996)
- fix panel data bugs: observation strings were broken for
frequency greater than 99, and the "panel" command did not
obey the "cross-section" option (thanks to Mihai Nica for
the report)
- frequency plot, against gamma distribution: add Locke's
nonparametric test for the null hypothesis that the variable
in question is gamma
- add facility to generate cross-sectional unit dummies with
panel data ("genr unitdum")
- add facility for nesting loops and fix some loop bugs
- fix some bugs in LaTeX equation display
- fix some breakage in online help system (e.g. for boxplots)
- add support for Latin 2 characters in gnuplot graphs
- add support for compacting daily data to weekly or monthly
- add support for daily data with a 6-day week
- fix for SUR systems with differing numbers of regressors
across equations (thanks to Jack Lucchetti)
- improve convergence of GARCH function (Jack Lucchetti)
- make tobit command robust with respect to the scale of the
dependent variable
- add stack() function to genr command: creates a stacked
series from a comma-separated list of series, or a
range of series as in stack(x1..xn)
- add a genr "special": genr markers = ... can be used to
manipulate the observation marker strings -- details in
panel.pdf
- support "genr varname[obs] = ..." for setting individual
observations in a series
- Text/CSV data importer: handle calendar dates for non-daily
data; handle description appended after the data
- add "vif" command to calculate Variance Inflation Factors
for independent variables in a regression; also available
in the model window under Tests/collinearity
- switch to St Louis Fed database as gretl "sample" database
(the BCIH database is getting too out-of-date)
- add the Anscombe Quartet to the gretl data files
- add Mankiw, Romer, Weill (QJE, 1992) dataset and replication
script, also Nelson and Plosser dataset
- on Windows, make EMF graphs respect the user's choice of
plotting colors (required a modification to gnuplot)
- add option to save graphs in EMF format
- various minor gui fixes and enhancements
- update Windows version to glib 2.4.7 and gtk 2.4.13
- start adding user-defined functions
- add mechanism for customizing gretl's tex preamble
7/23/2004 Version 1.2.9 [Back to top]
- fixes for LaTeX printing of autoregressive models
- correct Spanish help strings
- fix font problem for some objects in session window
- fix for exporting time-series data to R version 1.9
- add option to export data to PcGive .dat format (thanks to
Michal Kurcewicz)
- enable TRAMO analysis of annual time-series data
- Windows version: ensure user directory gets created! This
was broken in gretl 1.2.6 and 1.2.7, which could produce
strange-seeming problems.
- hccm command: ensure that the F-stat shown for the model
is based on the robust covariance matrix; also ensure
the robust F-stat is used when robust standard errors are
selected with the ols command
- add start-up option to dump gretl configuration to a text file
('-c' or '--dump')
- small fixes for nonlinear least squares
- Replace fixed IP address for database server with hostname
- Fix for copying material from gretl using Ctrl-C on Windows
- Update Windows version to gtk 2.4.3 runtime
- add "restrict" command for general Wald test of linear
restrictions on a model
- add "ok(x)" as a short synonym for "!missing(x)" in the
context of the genr command
- change smpl behavior: restrictions now cumulate by default
- add "smpl" option: "--replace" makes a sample restriction
start from the full dataset rather than cumulating with any
previous restrictions (note: this was the old default)
- another new "smpl" option: "smpl 100 --random" selects 100
observations from the dataset at random
- allow the "--robust" option to the "var" command
- fix for effect of Up/Down keys in main gretl window
- gretlcli in batch mode: pass "#"-style comments through
to the output
- fix HCCME bug: used way too much memory with large data
sets
- add a new "set" option, "force_hc": if set to "on", then
the robust option to ols always produces plain HC standard
errors, not HAC (by default, HAC is used with time-series)
- profile and optimize some gretl_matrix functions
- allow specification of consecutive lags in a regressor list
using a shortcut of the form "x(-1 to -4)"
- fix some subtle genr bugs, and some breakage in the nls
command following earlier changes in genr
- small configure fix: try harder to find libpng if wanted
version 1.2.8: a "test-only" version, with some of the features
of 1.2.9, depending on the date of the build.
6/13/2004 Version 1.2.7 [Back to top]
- add Italian helpfiles (thanks to Cristian Rigamonti)
- update Spanish helpfiles
- add start-up option "-e" or "--english" to force use of
English strings when LANG would indicate translation
(also recognize environment variable GRETL_LANG: if this
is "english" or "C", run in English)
- allow for missing values (silent) in audio graphs
- add a few keystrokes to the main gretl window:
'a': audio (read list of variables)
'x': stop audio
'e': edit attributes of current variable
't': plot current variable
- make start-up tests for tramo and x12a less verbose
- fix bug in deletion of scalar variables from the dataset
- fix bug in built-in spreadsheet: clicking in a blank
column could crash gretl
- fix broken behavior of built-in spreadsheet for gtk
2.0.X versions
- fix various bug and typos in documentation
- build tweaks for OS X
- make available a pre-built disk image for OS X
5/25/2004 Version 1.2.6 [Back to top]
- add Hodrick-Prescott filter, as "hpfilt" function in the
genr command
- revamp xls importer: should now be substantially more
robust
- begin adding experimental audio graph feature for visually
impaired users
- improve handling of non-ASCII characters in data files
- fixes for Spanish help files
- more fixes for build process
- GNU Octave data export: write all data as single matrix
- fix printing of "periods" in periodogram output (thanks to
Tadeusz Kufel)
- fix generation of periodic dummy variables for daily data
(again thanks to Tadeusz Kufel)
- with gtk >= 2.4.0 on Linux, use the new gtk file chooser
- update the Windows gtk version to 2.4.1
5/4/2004 Version 1.2.5 [Back to top]
- fix problem with acf and pacf calculations
- add Prais-Winsten estimator for AR1 models ("pwe" command)
- fix bugs in gretl "graph page"
- fix some translation bugs
- fix headers for compilation on Mac OS X
- fix bug in lmtest option handling
- make Excel importer a little more robust
- bug fix for saving data files with very large number
of variables (tested with 548)
- bug fix for "diff" and "ldiff" commands with panel data in the
form of stacked cross-sections
- plain text output: give option of copying as RTF (puts the
output into Courier New at 9 points)
- in gui, disallow deleting variables when sub-sampled
- setobs command: be more flexible in regard to the format of
the "startobs" field
- saving text files: try to improve handling of CR/LF
- minor fixes to model printing
- gnuplot fix: make copy-and-paste into Microsoft applications
work again (this was broken by an MS security update of April
13, 2004)
3/30/2004 Version 1.2.4 [Back to top]
- Add ARMAX models: ARMA with exogenous variables
- Add GARCH models, using Fiorentini-Calzolari-Panattoni approach
(Journal of Applied Econometrics, 1996)
- Add "resample" function to genr, for bootstrapping test
statistics
- Loops: allow use of a variable name rather than a number to
control a "number of times" loop
- Fix genr bug: problem with an undefined varname in a lag
context
- Fix bug: when saved via the objectname mechanism, arma models
were appearing both as saved models and as saved text
- Fix bug: autocorrelation test from model window Tests menu
was broken by new options mechanism
- Fix bug: "--verbose" option for Johansen test was broken
- Fix curious Windows bug: line-endings in scripts were not
recognized in some cases
- Improve path-searching for data files
- Beef up error-checking on opening of native gretl data files
- Improve error-checking on pvalue command
- Fix for Excel importer: allow more than 16K rows
- Error-checking for new varname in genr
- Add robust standard errors/VCV for tsls command
- Allow use of wls command in a loop
- CSV import: add automatic conversion of string variables
to code numbers
- Add Three-Stage Least Squares (system type=3sls): this is
still at the testing stage
2/24/2004 Version 1.2.3 [Back to top]
- Add Tobit model command, with a new plugin (thanks to Jack
Lucchetti)
- Add color selector for the histogram in frequency plots
- Allow appending/merging of gretl-format data in GUI program
- Fix bug: genr was broken for some special cases of
repeated function calls
- Fix bug: normal pvalues were being shown as zero for
large |z-scores| > 9.0
- Fix bug: the NIST checker plugin was broken for locales
that use ',' as decimal point
- Daily data: construct a special plotting variable so
daily time-series can be graphed better; also allow
for 4-digit years in daily dates (YYYY/MM/DD) on input;
also fix "genr" so that dummies can be constructed
as in "genr dum = t < 1987:10:19" (note that the
slashes must be replaced by colons in this context)
- Text objects: in scripts and the gui console, many
commands can now have their output diverted to a saved
object (e.g. "ADF <- adf 1 Ct"). As with models and
graphs, these objects appear in the Session icon
window, and they can be displayed and freed using
objectname.show, objectname.free
- Make the model table available via script and gui console
(new command "modeltab")
- Add "graph page" in session view: allows the user to
create a page (which may then be viewed and printed)
containing up to eight graphs (requires latex, dvips and
a postscript viewer)
- Add an OLS option, "-r" or "--robust", which calls for
the use of White's robust standard errors (HC0) in the
case of cross-sectional data, or HAC in the case of time
series data. Details can be configured using the new
"set" command.
- New command, "set", for configuring some program
parameters via script or console
- Fix bug: latent problems when adding an observation to
the working dataset
- Fix bug: set the frequency and starting observation could
fail when the local decimal point is not '.'
- Fix: make the genr functions "coeff()", "stderr()" and
"vcv()" work to retrieve statistics from ARMA models
- Fix bug: when appending CSV data to an existing dataset,
could get uninitialized values at the end of the series
- Allow backslash-continuation of long commands in the gui
console
- Updated help files and manual
2/2/2004 Version 1.2.2 [Back to top]
- Make AIC retrievable via genr ("genr foo = $aic")
- Fix brokenness in RTF printing/copying of ARMA models
- Add calculation of modulus and frequency for ARMA roots,
and add printing of these to LaTeX and RTF output
- Fix bug in accessing critical values of F (Utilities menu)
- Add DFFITS values to the "influential observations" model
test; also expand the help entry for this
- Add a little explanation of Johansen cointegration test
results to GUI helpfile
- Fix bug: when you go to save data imported from gnumeric,
the suggested file suffix should be changed to ".gdt"
- Fix bug: when importing data from Excel with a non-zero
row offset, respect the offset when checking for date
strings
- Fix bug: adding a variable to the built-in spreadsheet
(GTK 2 version) when there are already 6 variables shown
- Fix bug: handle quotation marks within descriptions of
variables (XML encoding)
- Fix bug: bad behavior when you put an Excel file to open
on the gretl command line, then cancel the import
- Fix bug: X-Y plot with "factor separation" went wrong in
some cases (spotted by Ignacio Diaz-Emparanza)
- Fix bug: weighted least squares with a dummy variable for
the weight crashed gretl when using QR decomposition
- Fix bug: a syntactically incorrect "genr" expression could
lead to a crash (thanks to Steve Walesch for spotting this)
- Fix bug: set panel structure dialog very messed up in the
gtk-1.2 version of the GUI program
- Fix bug: Sorting variables by ID in the main window did not
work correctly for more than 9 variables
1/22/2004 Version 1.2.1 [Back to top]
- Add Cristian Rigamonti's Italian translation
- More flexible and general support for collections of data
files or script files
- Add support for the data files from Davidson & MacKinnon,
Econometric Theory and Methods (ETM)
- Make the NIST linear regression test suite available via a
plugin
- Add ARMA model to the variable menu (plus "arma" command)
(thanks to Jack Lucchetti)
- Fix bugs with displaying a TSLS model in the GUI program
- Fix bug in display of actual, fitted and residuals in GUI
program
- Fix occasional bug in copying data to clipboard as CSV
- Add a mechanism for configuring gretl as a networked
application under Windows
- Small fixes for data importation
- Don't overwrite CSV datafiles with gretl format data
- Add a logistic model command and menu item
- Fixes for reading of recent RATS databases (thanks to John
Frain for a useful bug report)
- Some updates to the manual and help files
- Enhanced LaTeX equation output (thanks to Alan Isaac for the
suggestion)
- Fixes for several nameless bugs
- Various GUI tweaks
12/14/2003 Version 1.2.0 [Back to top]
- add "rename" command (rename a variable)
- add "printf" command for generating formatted output
- "delete" command: add the ability to delete specified
variables, identified by name or number
- add "eval" command: syntax as in a "genr" formula; the
result of calculation (or retrieval) is printed then
discarded, not added to the data set as with genr
- Replace the internal workings of "genr" so that it can
calculate dynamically (e.g. "genr y = .6*y(-1) + u"),
which in effect makes the "sim" command redundant
- Make principal components analysis available via the GUI
- Improve formatting of descriptive statistics output when
only one variable is selected
- Allow adding of "influence" values (calculated by the
"leverage" test) to the data set in the GUI program
- When adding a variable to the dataset from a model window,
give the user a chance to edit the name and description
- Windows: add option to emulate Windows look (via WIMP)
- add "--no-restore-data" to R invocation
- Fix for launching R from gretl on Windows XP
- Improve creation of time-series datasets for use in R
- Fix problems created by tabs in the command line
- Two-stage least squares: add list of instruments to printout
of model results; fix crashing bug in printing covariance
matrix of coefficient estimates
- Improve the mechanism for detecting the presence of a lagged
dependent variable in a regression list; also show the
Durbin-Watson stat in addition to Durbin's h when a
regression includes a lagged dependent variable
- Fix bug in graph editing for non-English locales
- genr command: add facility for using date strings in boolean
comparisons, e.g. "genr dummy = t > 1973:4" will work for
quarterly data
- Allow '#' comments in loop constructions
- sim command: allow omission of starting and ending observations,
to simulate over the full sample range; also allow the use
of a minus sign to get the negative of a variable
- VARs: add facility to separate deterministic terms such as
trend and dummies from the other variables in the regression
list; add Durbin-Watson statistic to model printouts; add
impulse responses and variance decompositions
- Improvements to data-editing spreadsheet in gtk-2.0 version
- New, improved dialogs for setting the sample range and sub-
sampling using a dummy variable
- Add color selection to the graph editing dialog and improve
the mechanism for selecting a graph font (only applies if
gnuplot has the "new" PNG driver, i.e. gnuplot 3.8)
- Graphs: add option of removing the top and right border lines,
if the graph does not have a "y2" axis
- minor fixes to the build process
- updates to gtk-1.2 GUI (make more consistent with the gtk-2.0
version)
- Allow for renaming of saved session models and graphs in gtk2
version of program
- Add a "lite" version of gtksourceview to the gtk2 source package
- Add recognition of dated daily data in importing from ASCII:
dates must be like 89/12/31, i.e. YY/MM/DD
- Loops governed by a convergence criterion: set a failsafe
maximum number of iterations, namely 250. This can be altered
by setting the environment variable GRETL_MAX_ITER
- Update to gcc-3.3.1 and mingw-runtime-3.2 for Windows build
10/31/2003 Version 1.1.6 [Back to top]
- Test release for Fedora (Linux) packaging
10/24/2003 Version 1.1.5 [Back to top]
- Updates to manual
- Gtk2 version: Make selection dialog tall enough for wls,
tsls and ar model selections
- Model window, graph menu: if there are two independent vars,
offer the option of a 3D fitted/actual plot
- Set sample range: in this dialog, give as default the current
sample range, not the full data range
- When a session uses data imported from a database, be smarter
on saving the session: ensure the script includes a command
to open a datafile (get this saved first if need be)
- Database browser window: close this when a database is
selected
- GUI model windows: allow drawing of graphs relating to model
data, even if the sample has been changed since the model
was estimated
- Model table: make a LaTeX preview available
- Windows installation: fix typo in name of uninstaller program
in the Programs menu
- Windows version: bug fixes and enhancements for gretl_updater,
and add a simple logging mechanism to aid in trouble-shooting
- Windows: add the ability to print a gnuplot graph directly
- ols command: add '-q' (quiet) option to suppress printing of
results (and add notice of this to gretlcli.hlp)
- Consolidate "webget" code -- one version for all uses
- gtk2 version: replace "system()" calls with g_spawn()
- Add model selection statistics to output of adf command
- Improve recognition of dated data when importing from Excel
or Gnumeric
- Add an error message for an attempt to run TRAMO when the
sample size is > 600
- Fixes for reading of plot range info in graph window in
non-English locales
10/6/2003 Version 1.1.4 [Back to top]
- Make auto-entry of varname for sampling with dummy smarter
- Fixes for display of data values with huge datasets
- Fix to handle "weird" cases in plotting frequency distributions
- Auto-detect gnuplot's ability to use "filledcurve" style
- Translate a few untranslated strings in gtk-1.2 version
- Graph labels: make 3 labels configurable rather than just 2
- Editing gnuplot commands: fix for accented characters
- Reading CSV files: fix bug in parser for date strings
- NLS: ensure messages printed to the console are not in utf8
- Loosen up a little on "Appending" data: allow for the situation
where the sample range for the added data is not identical
to the original range
- Fixes for printout of fitted, actual and residual, and for
forecasts with standard errors, in the case of missing data
- Fix to gtk2 data spreadsheet (quell warning on appending an
observation)
- Data spreadsheet: give a warning on close if there are unsaved
changes
- Improvements to printing of data values in certain numerical
ranges
- Import of CSV or plain text data: if no variable names are
given, construct them automatically ("v1", "v2", etc.)
- 3-D plot command: fix for localized decimal point, and use
rxvt to control gnuplot if xterm is not available
10/1/2003 Version 1.1.3 [Back to top]
- Fix crashing bug in viewing dataset info
- Improve formatting of "influential observations" test output
for non-English locales
- Correct the sign of the fractional integration parameter
estimate in the periodogram command ("pergm")
- Give first-time gretl users a chance to select a gretl
"user directory" (with ~/gretl being the default)
- Fix problem with running a script in the gretl console, when
the script begins with comments plus blank lines
- Brighten up some menus with stock icons (gtk2 only)
- Use gtksourceview to do syntax highlighting for scripts
(gtk2 version only)
- Change function of main-window toolbar edit icon: from launch
an external editor to start a new script
- Fix bug in meantest command: the -o flag was doing the
opposite of what was stated in the manual
- Fix bug in built-in spreadsheet in re. adding variables or
observations
- Ensure that the listing of variables in the main window
is updated after running a script
- Add facility to exclude named observations with smpl command
(e.g. smpl -r obs!="USA")
- Add "nobs()" function in genr: gives the number of valid
(non-missing) observations for a data series
- Improve placement of observation labels on graphs
- Improvements to gnumeric data importer
- Add facility for generating 3-D data plots
- Add "import ASCII" data option
9/9/3 Version 1.1.2 [Back to top]
- Fix bug in time-series plot from popup menu in main window
- Fix build problem: "make install" could fail to install the
GUI program under some conditions (thanks to Dirk Eddelbuettel
for spotting this)
- Update the Spanish translation
- Win32 version: switch to the native Windows font selection
dialog box (some problems were reported with the modified
gtk font selection on win32)
- Win32 version: update to gtk 2.2.4 and pango 1.2.5
- Library: switch to using the Cephes gamma() code throughout,
and delete some redundant code
9/7/3 Version 1.1.1 [Back to top]
- Graph editing: add facility to write labels onto the graph
- Graph editing: allow removal of auto-fitted OLS line
- Graph editing: allow choice of TrueType font if gnuplot supports
this option (requires libgd and freetype)
- Add compaction method (e.g. average, sum) as an attribute of
variables -- settable under Variable/Edit attributes
- Add "outfile" command to send output of selected commands to
a specified file
- In importing data from a database, recognize if a variable
of the same name is already in the working dataset
- Fix memory bug in opening remote database
- Start adding Principal Components Analysis
- Fix "pmax" bug in printing of actual, fitted and residuals
- When using "<-" save mechanism for models, preserve the model
info so it's accessible for genr, etc.
- Fix bug in numbering of auto-generated lagged variables
- Fix bug in parsing of gnuplot { ... } special command list
- Fix bug in model table (wrong coefficient numbering)
- Fix bug: when a file is dragged onto gretl, opening the
file failed if it has a space in its name
- configure: require correct version of GMP (>= 4.0.1)
- Fix bug: gretlcli.exe on win32 didn't find user directory
- Workaround: command-line invocation of gretlw32.exe with
datafile could crash (I think this was a Windows bug)
- Fixes for forecast-with-errors graphs
- Fixes for "influential observations" test calculations, in
extreme cases
- Windows version: update to gtk 2.2.3 runtime
7/23/3 Version 1.1.0 [Back to top]
- Extend DATAINFO structure -- allow specification of "display
names" for variables (these are shown in graphs)
- Provide "label" command for setting the description and/or
display name for a variable
- Add script commands for saving and displaying models and
graphs
- Fix effect of sample range on the "genr" command
- Improvements to the gretl console (command history, etc.)
- Add the Johansen cointegration test to the gui menus
- Add option to use QR decomposition instead of Cholesky for
calculating regression results
- Fix bug: periodogram command failed in batch mode
- Make uniform() PRNG generate in the standard range 0-1
(rather than 1-100, as it was previously)
- Small modifications to path-searching
- Updates to the Spanish help files
- Activate a few missing translated strings
- Fix "sim" command so it can be used to modify just one
observation in a given series
- Fix bugs associated with starting a dataset by importing from
a database
- Allow access to RATS databases via command line and scripts
- Fixes (I think) for R startup
- Clean-up: remove all wasted array elements
- Update to libtool 1.5
- Update win32 version to gtk version 2.2.2
5/26/3 Version 1.0.9 [Back to top]
- Add non-linear least squares ("nls" command, Levenberg-
Marquandt algorithm from minpack, with either analytical or
numerical derivatives)
- Fix focus bug when editing the description for a variable in
the gretl main window
- Add a new means of grabbing fitted values from a regression:
"genr newname = yhat" (parallels "uhat")
- Updates to manual (includes details on new nls command)
- Spanish version: provide access to English help files as well
as Spanish
- Add rudimentary scripting capacity to gnuplot command (can
append "{ ... }" block to gretl gnuplot command, containing
literal gnuplot command lines, each terminated with ";")
5/2/3 Version 1.0.8 [Back to top]
- Require lapack
- Fixes to build for Mac OS X
- Avoid duplication of "against time" in model graph menus, when
time is included as a regressor
- Fix bugs in Find dialog
- Correct function that detects whether or not series should be
scaled for time series graphing (case of negative values)
- Fix bug in boxplots with boolean conditions attached
- Ensure all plot windows are not resizable
- Add "leverage" command (uses lapack) to detect leverage points
and influential observations for a given model
- Fix "diff" and "ldiff" commands for panel data sets (i.e.
show missing values as missing)
- Fix bug in editing of varnames and labels (could get onto
the wrong line in some cases)
- Fix gretl's startup routine for Gnu R so that it never over-
writes the file .Rprofile
- Enhancements to Excel importer plugin (better warnings)
- Permit opening Excel or Gnumeric data from initial command-
line invocation of gretl
- Autocorrelation test: add calculation and display of serial
correlation-robust standard errors, if significant at the 5
per cent level
- Fix brokenness in saving boxplots to session file
- Refinements to CSV import function -- handle space-separated
data with multiple/variable spaces between columns
- Fix bug in range-mean graph when there are missing values at
the start of the data range
- Fix bug in pacf calculation and plot for long lag lengths
- Add experimental "model table" in session icon view (not submitted
for translation yet)
4/9/3 Version 1.0.7 [Back to top]
- Correct the attribution of autocorrelation test (Breusch-Godfrey
rather than Breusch-Pagan)
- Fix bug in display of remote database info line (Gtk-2.0 version)
- Use a finer step-size for Hildreth-Lu search
- In GUI program, display Hildreth-Lu plot using gnuplot
- Translate a few more untranslated strings
- Fix bug in title of residual graph
- When the user installs a local copy of a remote database,
give the option of opening it right away
- Add "coeffsum" command: gives the sum of coefficients, and
standard error of the sum, for selected independent variables
in a given model
- Modify the White's test function, so that it doesn't fall over
when a variable and its square are among the original regressors
- Fix drag and drop of data files and script files from desktop
or file manager onto the main gretl window
- Add drag and drop of data series from a database window into
the main gretl window (i.e. into the current data set)
- Add general time-series data compaction function (under Sample
menu)
- Fix various memory-usage bugs revealed by port to Mac OS X
- Fix installation of bcih database so that it works on big-
endian machines too
3/16/3 Version 1.0.6 [Back to top]
- minor adjustment to model print output formatting
- fix for correlogram graph (was incompatible with older gnuplot)
- fix a couple of minor gretlcli bugs
- add (optional) LAPACK support for various "advanced"
features
- fix time-series attributes of data9-1, data9-10
- prevent normal random generator from outputting NaNs
- fix bug with use of "store" command in a loop, in GUI program
- fix translation problems that disabled some menus when not
shown in English
- fix translation problem that caused a crash when trying to save
data under a different name in non-English versions on Linux
- add new commands, "system" (supports SUR) and "coint2"
(Johansen trace test for cointegration)
- add robust LM test for use with adding/omitting variables in
relation to a model estimated using hccm
- permit comments starting with '#' on first line of script
- change to Mersenne Twister for gretl's PRNG, and supply the
Marsaglia test suite for the PRNG in the gretl source package
- make correlogram and periodogram functions more robust in
the face of weird data
2/20/3 Version 1.0.5 [Back to top]
- fix missing translations; fix build of rpm
- add support for Gujarati textbook data files
2/19/3 Version 1.0.4 [Back to top]
- add sst() function in genr for Total Sum of Squares of a
variable
- fix bug in genr: a long genr command could overflow the space
allocated for variable labels
- fix bugs in graphing: "non-standard" graphs could cause problems
including crashes
- fix bug in coefficient printing: sometimes values were being
truncated at 5 figures
- when dataset is sub-sampled, add dialog offering option to
restore full data range before saving data
- add recognition of environment variables GRETL_PNG_GRAPH_FONT
and GRETL_PNG_GRAPH_FONT_SIZE (e.g. "verdana" and "8") for
generation of PNG graphs when gnuplot is linked to gd, and
gd has TrueType font support enabled
- The sample data files selector and practice scripts selector
now offer a tabbed collection of the available sets of files,
and the associated menu items are simplified
- Replace Box-Pierce statistic with Ljung-Box throughout, and
revamp the correlogram graph
- Add several configuration options to the tramo/seats plugin
- update libtool to 1.4.3 and autoreconf
- raise library revision number to 4
2/4/3 Version 1.0.3 [Back to top]
- Add PNG graphs to win32 version, and standardize on PNG graphs for
all display including graph editing
- Add facility to identify labeled data-points in a scatter plot
- further fix to pathnames with .Rprofile for win32
- LaTeX view: fix for dvi file living in a directory with
spaces in its name on win32
- Add option to copy as LaTeX or RTF for the various "Model data"
windows (e.g. forecasts with standard errors)
- Reorganize management of win32 registry entries
- Add tramo/seats for win32 version, and make tramo support the
default for Linux
- In window displaying a single data, add menu options to show
the series sorted, to graph the variable, and to format the data
values to a specified number of significant digits or decimal places
- Fix bug in command echo for "store" and a few others
- Add selector for variables to be saved when doing "Save as"
or "Export"
- Add facility to browse for programs under the Preferences,
General, Programs tab
- Add facility to copy data to clipboard as CSV
- Update win32 version to gtk version 2.2.0
- Bring various details of the gtk-1.2 build in line with the
gtk-2.0 version
- Add configuration option to avoid libreadline dependency
("--without-readline") and build the gretl rpm without readline
- Updates to manual
- Raise shared library revision number to 3
12/9/2 Version 1.0.2 [Back to top]
- generalize support for X-12-ARIMA to gtk-1.2 version, and
win32 version (TRAMO/SEATS proving is more difficult)
- make x-12-arima support the configure default, if the x12a
binary is found
- add installation script for tramo program (in utils)
- make range-mean graph editable via gui
- fix utf8/character encoding issue with editing gnuplot
commands and sending the commands to gnuplot
- fix bug in test for balanced panel and reading of panel
structure
- panel models: implement a valid autocorrelation test, and
make the ARCH test unavailable
- fix bug in generating lagged variables with panel data
- put a limit of 1024 on the number of forecasts that can be
called for, in the "forecast with standard errors" routine
(could exhaust virtual memory)
- add menu item to restructure a panel data set from stacked
cross-section to stacked time series
- create data set using internal spreadsheet: fix bug in
the setting of date strings
- fix bug in invocation of GNU R under Windows (needed to
replace "\" with "/" in paths for use by Rgui.exe)
11/27/2 Version 1.0.1 [Back to top]
- fix to configure script in re. glib-2.0
- fix missing name of dependent var in gui mpols printout
- remove RTF and LaTeX items from mpols copy options (not yet
implemented) in gui2
- fix bug in saving gnuplot commands to file
- updates to updater function in gui program (both gui and gui2)
- add rudimentary plugin support for TRAMO and X-12-ARIMA
(experimental, and disabled by default)
- fix font encoding for postscript graphs in gtk-1.2 version of
gretl_x11 (also make postscript graphs in color by default)
- fix bug in automatic generation of lags for regression (the
command was not getting recorded properly in the log file)
- fix bug in model printout when F() is undefined (line-break
was missing)
- fix bug in LaTeX output when a coefficient equals zero
- make "sim" command available in loops (and update docs)
- raise "revision" number of *nix library
11/15/02 version 1.0
- change *nix library name to libgretl-1.0
- improve sizing of some windows
- incorporate new translations
- update appearance of gtk-1.2 version of gui
- fix the stats available in connection with a LAD model
- open the champagne (hope this is not premature!)
11/06/02 version 1.0pre4 (for translators)
- fix bug in progress bar when loading large data set
- fix bad "smpl -r" bug in gretlcli
- add Ramsey's RESET test (command "reset")
- improve consistency of editing commands
- more translation fixes
- small LaTeX output fixes
- fix gnuplot PNG output for gnuplot version 3.8
- fix encoding for gnuplot postscript output under NLS
- fix pango encoding issue for gnuplot axis label strings
- fix gnuplot "keyspec" bug
- add auto-scaling of gui windows if the user has selected a
large font
- add printing of rho (used in quasi-differencing) to LaTeX
output for corc and hilu models
- enhanced support for Wooldridge data files
- add "-t" flag to print command: print variables to ten
significant digits
- improve saving of scalar variables, and documentation of same
- updates to full manual
- make names and labels for monthly and quarterly dummies more
descriptive
10/10/02 version 1.0pre3
- improve accuracy of calculation of standard errors, and
numerical accuracy on MS Windows in general
- make gretl databases installable in user directory as well as
system database directory
- make console output scroll to cursor in gtk2 version
- fix more translation issues
- fix bug in normality chi-square test for large n
- fix bug that prevented CUSUM plot from displaying
- fix normality of residual test: print to model window was
incomplete
- fix translation issue in LaTeX print of equation
- backport fix for impulse plot to gtk-1.2 version
- fix bug at exit of gretlcli with "quit" and no filename
- fixes for bugs in gnumeric file reader
- fixes for Excel reader -- handle missing obs and Far East info
- fix for gretlcli under NLS: ensure that the translations come
out in the local encoding, not utf-8
- remove Gtk 2.0 right click menu on text windows
- update translations and integrate the French translation
09/20/2002 version 1.0pre2 (pre1 was "for translators only")
- add range-mean graph option to GUI Variable menu, and
as new command "rmplot"
- add retrievable variable $pd = data periodicity
- add min() and max() functions to the functions available
for use with the "genr" command
- add Least Absolute Deviations regression ("lad" command)
- fix bugs in saving output to named file, in both gretlcli and
gui program
- fix bug in boxplots for languages other than English
- fix error in df in sample script ps4-2.inp
- fix translation of date strings in gretl output
- linux: move plugins to ${prefix}/lib/gretl
- gnome 2.0: use gconf apparatus
- gtk 2.0: fix remove/add toolbar breakage; change name of program
back to "gretl" and "gretl_x11" (not "gretl2", "gretl2_x11")
- revamp and internationalize TeX and RTF printing functions
including those for summary stats and correlation matrix
- make TeX format available for all model windows
- remove HTML print/copy options: can't maintain four formats
with internationalization!
08/25/2002 version 0.999
- add support for gtk 2.0 and gnome 2.0
- enhanced gui with gtk 2.0
- add "make check" target to test against the NIST reference
data files for numerical accuracy
- improve support for building on win32
- bug fixes in boxplot functions
- always show 6 significant figures for coefficient estimates
and standard errors in regression output
- numerous NLS/translation fixes, improvements
- NLS translation of TeX model results
- bug fixes for Excel importer
- reorganization of gui source files
- fix bug in random variable generation in gui
- add support for data files from Jeffrey Wooldridge's
econometrics textbook
- scrap ugly overloading of "clear_model" function
- fix miscellaneous bugs revealed by valgrind
- raise library version to 16.0.0
07/09/2002 version 0.998
- increment library version to 15.0.0
- update Spanish translations
- fixes to RPM build process
- switch to more accurate (cephes) algorithms for p-values
- add materials for building gretl on win32
- try to support LC_NUMERIC conversion automagically
- add option to Preferences to enable/disable localized
LC_NUMERIC
- fix listings of recently opened files on menus under NLS
- fix up formatting of fitted/residual printout
- activate translations of "session" popup menus
- extend LaTeX menu for models with choice of saving as full
document or tex fragment for inclusion in document
- correct over-eager rejection of some variable names in data
files ("c" and "C")
- recompile gretl_updater.exe with new zlib dll
- fixes to graph editing (this got broken by NLS changes)
05/28/2002 version 0.997
- increment library version to 14.0.1
- NLS fixes: disable "foreign" LC_NUMERIC for the present;
gnuplot graphs should work now; some missing
translations are activated; minor formatting improvements
- screen scalars out of variable list in graph selection dialogs
- ensure appropriate headers are included in workbook.c
- enable daily data in gretl databases
- fix bug in daily calendar (printed dates wrong at end of
some years)
- portability fixes to configure script
05/24/2002 version 0.996
- Raise library version to 14.0.0
- closing a session clears the selector dialog presets
- fix bug with scalar data and the data-editing spreadsheet
- revert boxplots to the older selection dialog (so user can add
boolean conditions)
- Regressions: print warning when data matrix is close to
singularity
- add Gnu Multiple Precision plugin for running multiple
precision OLS regressions (unix only, so far)
- Add Spanish translations and make enabling NLS the default
- Try to fix the win32 GUI font for high-res monitors
- tweaks to the configure script
- updates to the help files and manual
05/15/2002 version 0.995
- Raise library version to 13.0.0
- numerous small updates to facilitate internationalization
- give the user a choice of delimiters (comma, space or tab)
when importing or exporting "CSV" data
- strengthen checks on invalid variable names in data files
- improve recognition of scalars versus vectors in genr
- fix bug in Create Dataset: undated data appeared as "annual"
- add descriptive report on dataset to Data menu
- fix bug in writing Utilities/R session: R data set was being
written in wrong (CSV) format
- fix bug: datafile name was not being cleared on Close Session
- refinements to Makefiles: don't expand CFLAGS multiple times
- handle spaces in filenames when stacking the "open" command
- fix bugs with editing gnuplot graphs
- get win32 LaTeX view working properly
- bugfixes for gnumeric and excel spreadsheet importers
- bugfix: unitialized data in cointegration test with constant
- add new, improved dialog box for specifying models, etc.
- remove all system() calls on win32
- switch to a more "standard" (I think) presentation of WLS
statistics
- fix some subtle bugs with the genr() function (scalars versus
vectors in conjunction with sub-sampled data)
- add checks for illegal variable names when importing from
CSV, Excel and Gnumeric
04/21/2002 version 0.994
- fix bugs in win32 copy to clipboard
- fix RTF printing of correlation matrix
- fix bug in genr (generating a scalar when data set is sub-
sampled)
- fix bug: editing periodogram graph using GUI
- partially update docs
- distinguish between $T (number of obs used in last model) and
$nobs (number of obs in current sample range) in genr
expressions
- fix bug in gnuplot PNG display for remote X sessions
- add configure option "--disable-png-graphics" to revert to
old-style behaviour of gnuplot in case of problems
- remove upwanted dependency on libgcc_s in Red Hat binary rpm
- update versions of gtk and glib for win32
- fix build problems revealed by the Debian package build
04/15/2002 version 0.993
- add option to specify lag order in autocorrelation test;
also add Box-Pierce and Ljung-Box stats to test output
- fix brokenness in Excel importer on win32
- fixes to setmiss command (console and script modes)
- enable pop-up menu for all gnuplot graphs under Linux/unix;
also add facility to print graphs directly under gnome,
and to zoom graphs on X11 display
- make DVI viewer configurable and make "latex view" functions
work on win32
- add inclusive inequality operators '>=' and '<=' to genr
- fix bug in genr: operator priority was not being fully respected
in all cases
- fix bug in Replace function for script edit window
- fix bug that crept in, in reading old-format data files
- fix bug: possible loss of precision when writing data file
- fix bug in statistical reportage with weighted least squares,
when the weight is not a dummy, but has some zero values
- add if, else, endif flow control for scripting
- add noecho command to suppress echo of commands in script
output
- modify "print" command to support printing of string literals
- add menu item to find a variable in the main window
- improve context help in script windows
- extend LaTeX printing of models to Cochrane-Orcutt and Hildreth-
Lu estimators
- add edit/print taskbar to more windows
04/01/2002 version 0.992
- bounds checking for critical values look-up; also add this
as a regular gretlcli command, "critical"
- add sanity checks for regression commands and ADF test
- add importer plugins for Gnumeric and Excel workbooks
- add "Append data" menu item
- fix bug in "Create data set" functions
03/20/2002 version 0.991
- fix several "out of bounds" bugs kindly reported by Tavis Barr
- streamline the online help system
- in GUI, allow deletion of any variable (popup menu)
- go to a two-dimensional data array
- get the data status indicator working better when the nulldata
command is used
- add menu items to set missing value codes for a particular
series or the entire data set, and add "setmiss" command
- changing or clearing the data set now closes all model and
stats windows (prevents crashes and garbage)
- add scalar datatype
- add progress indicator for reading/writing large data files
- add a refresh to the remote databases window after down-
loading and installing a database
- add handling for daily data (probably needs more work)
- update to gnuplot 3.7.2 in win32 distribution
- prepare the program sources for internationalization via
GNU gettext; add a blank .pot file to the source package
02/27/2002 version 0.99
- add option for HTTP proxy when contacting gretl data server
- fix bugs in drag and drop under Windows
- fix crashing bug when trying to save a data file with over
(about) 120 variables
- fix bug in editing new data header info
- add mechanism to record sample when estimating a model, so
that subsequent tests on that model can be referred to the same
sample
- add "native" printing of window contents under win32 and
gnome
- enhance boxplots command, allowing for boolean expressions
to limit the sample for individual plots in a comparative
group of plots
- add editing toolbar to editable windows with search/replace
and undo
- add context help (click on a command and press the "?" button)
for the edit command script window
- correct typos in supplied data files data3-2 and data8-3
- add "critical" command (gretl console and GUI script mode only),
to print critical values for a selected distribution
- add pvalue() function to the set of options available under
the genr command
- improve handling of scripts opened in the GUI (including a
prompt to save before running if changes have been made)
02/11/2002 version 0.98
- fig bug affecting frequency distributions and the test for
normality of residuals
- improve functionality relating to saving of data files
- fix some build problems
- add more functionality to boxplots
- several fixes and enhancements to "session" mechanism
- better handling of empty data header info
- substantial code cleanups in the interest of adding API
documentation
- allow copying LaTeX equation representation of models
to clipboard; also improve the typsetting of these
- change default user directory to ~/gretl under unix
- add option to make cwd rather than userdir the default
for file opening and saving
- Change the default data format to XML (optionally gzipped)
but retain backward compatibility with old ascii data
format
- add dependency on libxml (dll suplied for win32)
- switch from ".dat" to ".gdt" as default filename extension
for gretl datafiles (but add option to GUI to use ".dat")
- changing to storing data header info in buffer rather than using
repeated file I/O
- add search/replace dialog to window for editing scripts; also
Edit/Paste menu choice and choice of fixed font
- add drag and drop support for datafiles under win32,
register the .gdt suffix for gretl datafiles, and add an icon
for these files
12/14/2001 version 0.97
- fix breakage of var command in GUI
- add facility to draw boxplots
- add various formats to copy of models to clipboard (HTML,
RTF, LaTeX)
- enable LaTeX and RTF copying for correlation matrix and
summary statistics windows
- Add the LMF statistic (Kiviet, 1986) for the autocorrelation
LM test
- Add prompt to save data on exit, if the data set has been
changed or newly constructed
- Fixes to BOX1 datafile reading function
- small fixes to display of databases
- revert behavior in case of taking logs of series with non-
positive values (i.e. fail gracefully)
- add initial gnome support
- convert manual to SGML (DocBook) and update
- support directory names with spaces (yuck) under win32
- under win32, use registry instead of system-wide config file
libgretl.cfg and per-user gretl.rc; also use native win32
dialog boxes for more functions
- add check for GRETL_HOME environment variable
- make gnuplot command configurable under Preferences (meant
for win32)
- add a menu bar to the spreadsheet window
07/26/2001 version 0.96
- update to gtkextra version 0.99.16
- fix bug in Preferences/Toolbar entries
07/17/2001 version 0.95
- update to gtkextra version 0.99.15
- fix crash bug on opening a saved "session"
- fix bug that broke the "run" command on win32
06/14/2001 version 0.94
- make "false" the default for the automatic gretl update query
on startup
- catch errors in ar, tsls and scatters commands due to insufficient
arguments
05/11/2001 version 0.93
- fix problem with printing of actual, fitted and residual in the
case of a binary dependent variable
- CVS data export: put quote marks around the dates or observation
marker strings
04/23/2001 version 0.92
- library version goes to 6.0.0
- fixes to pprintf() function
- add "pooled OLS" model command and panel diagnostics test
(fixed effects and random effects models)
- various fixes suggested by lclint
- fix bug with zero return from, e.g., auxreg() when there's
an error and the return should be non-zero
- fix bug: logs command in script mode producing spurious error
- fix bug: when starting a new data set from a gretl database,
datainfo->markers was not set to 0, which could produce a crash
- add options for compacting data when adding a higher-frequency
series from a database to the working data set
04/03/2001 version 0.91
- library version goes to 5.0.0
- Disable saving anything other than native data file with
extension ".dat" (safety measure)
- Make backup of data file when saving under same name
- Further improvements/fixes to gnuplot graphing routines
- Add option to edit plot commands directly, for a graph
saved as a session icon
- GNU R: save cross-sectional data as a "dataframe"
- Remove the "with controls" gnuplot menu items: edit plots
via the session window instead
- Institute consistent file dialogs: using gtkextra under X,
and native Microsoft dialogs under Windows
- Bug fix in pprintf() function
03/28/01 version 0.90
- Raise version number to 0.90 reflecting fact that the feature
set for version 1.0 is now complete
- Raise library version to 4.0.0 due to changed interfaces
- Bugfixes and improvements for built-in spreadsheet
- Make more statistics accessible under model data menu
- try to auto-detect script versus data file, when given one
program argument and no flag
- add gretl_model_new() to lib
- cleanups in library code and headers
- enable frequency plot graphs in console mode
- add command-line options (for GUI program) to open a gretl
database on start-up.
- add "t" facility to genr (can say, e.g., "genr dum = t>1979.4"
for quarterly data, without previously generating t).
- add "for loop" facility with auto index variable
- auto-save changes to scripts in file viewer windows before
running
- TODO: manual is not fully updated yet
03/16/2001 version 0.70
- Changed a few library interfaces, and so raised the library
version number (now 3.0.0)
- update manual (please consult it on the new options below)
- add sub-sampling based on a boolean condition (smpl -r)
- make it possible to restore full sample after sub-sampling
based on a dummy variable or boolean condition
- add graphing option: scatter of Y against X, with the Y values
colored differently depending on value of a given dummy
- suppress echo of varnames in commands (e.g. summary) where
a full list of vars is constructed automatically
- improve mechanism for adding/deleting files from "recent files"
lists, and add list for command files
- improvements to numerical format in printing of data series
- fix a couple of graphing bugs (impulse plots not working,
frequency plot against gamma distribution wrong)
- fix precedence of boolean operators in genr expressions
("&" comes before "|", but both of these come after "<", ">"
and "="), and add "!=" operator
- mods to gtk items: disallow resizing of dialog boxes, make
clist titles passive
- fix "on exit" bug if several models estimated and not all
have associated saved commands
- fix memory bug when switching datasets
- fix issues with displaying data, with very large datasets
03/04/2001 version 0.69
- change one library interface to permit error box if gnuplot
command fails (and so raise library version number)
- bugfix: after running lmtest for non-linearity, the wrong
degrees of freedom were being shown, not in the auxiliary
regression window, but in the test summary appended to the
model for which the lmtest was run
- bugfix: with a genr command from the GUI, even if it was
erroneous it was being copied into the command log
- bugfix: parsing of genr expression without fully explicit
parentheses failed under some circumstances
- bugfix: multiple scatterplots did not display under MS Windows
- bugfix: doing add/omit in GUI did not update the model used
as the default for coeff() and such in genr commands
- console: improve/fix recall of mistyped command using the
up arrow key; respond to typing only at the prompt
- genr: add '&', '|' and '!' logical operators
- add program icon for X11
02/26/2001 version 0.68
- fix functions associated with opening datafiles
- fix brokenness in logging of commands to script
- "remember" models generated via GUI so that genr functions
such as coeff() work properly
- make Save/Export data menu items more compact
- Fix bug in (not) setting file extensions in some file "save"
actions
- Small improvements to "gretl console"
02/21/2001 version 0.67
- cleanups in datafile writing routines (avoid excessive
trailing zeros in data)
- make search in list boxes wrap
- fix bug: user configuration of toolbar had no effect
- fix bug: if a var is duplicated in list of regressors,
gretlcli crashed
- fix infelicity: if a var is duplicated in regressor list,
gui displayed blank model window along with error message
- fix bug: "gretl website" entry on toolbar under unix/Linux
worked only if netscape was already running
02/15/2001 version 0.66
- add string search to listing of Ramanathan data files
- add "plugin" for looking up critical values of statistics
- add a toolbar (partly configurable under Preferences)
- fix to deletion of auto-generated plotting vars
- add option to view confidence intervals for coefficients
(under Model data menu in model window)
- add the Ramanathan practice file ps2-1.inp
02/06/2001 version 0.65
- fix crashing bug in adding data to dataset from model window
- update bundled gtkextra to release 0.99.13
- various "internal" bugfixes and cleanups
- minor fixes to graphing routines, datafile saving
02/02/2001 version 0.64
- add tooltips to show full path of previously-opened files
- get "previous files" list working on win32
- minor improvements to file listing routines
- add (optional) sampling distribution graph to test calculator
- add -m (impulses) flag to gnuplot command
- fixes to auto-update mechanism (plus add documentation)
01/26/2001 version 0.63
- add gretl manpage
- add p-value finder to GUI
- add "hypothesis test calculator" to GUI
- fix error in normal batch p-value function
01/15/2001 version 0.62
- allow comment lines starting with '#' in input files
- allow estimation of a model with df = zero
- add CUSUM (parameter stability) test for OLS models
- fix to path reading for win32 (spaces in path)
- fixes to difference of means test, difference of variance test,
for situation where sample lengths are different
- small fixes to manual
12/09/2000 version 0.61
- fix to configure process for gtkextra
12/08/2000 version 0.60
- various fixes to distribution files
- change default user dir to $HOME/.gretl
- fixes to icon file selector
- basic link to GNU R
- further updates to manual
version 0.60pre
- revise PDF manual
- revise "scatters" command (multiple scatterplots) and add to gui
- add read/write of gzip-compressed data files (see store command)
- move descriptions of data files and practice scripts into
separate files (no longer hard-coded into gretl)
- add new data files from Ramu Ramanathan
- update to gtkextra version 0.99.12
- add installer program for win32
- fix context help bug
- renaming of various components (library is now libgretl, cli
program is now gretlcli, some directories have been renamed)
(note: version 0.50 was a Windows-only one, and was not really
released)
11/18/2000 version 0.40
- add critical value to printout of correlation matrix
- add capacity to read gretl databases from remote server
- add built-in data decompression with zlib
- rename exported library functions, avoiding leading underscore
- add libtool versioning
11/03/2000 version 0.32
- Add capacity to read BOX1-format data files
- Add lists of recently opened files to file menus
- fix various bugs connected with "session" concept
- fix gtkextra headers bug in make process
10/20/2000 version 0.31
- add gamma distribution to pvalue command
- fix some more bugs
- hush gtk warnings in win32 version
10/15/2000 version 0.30
- introduce experimental "session" idea
- update manual
- add "pergm" (periodogram) command. Also see "Spectrum" under
the GUI Variable menu
- add maxlag parameter to corrgm command; also add printing
of Box-Pierce Q stat, and partial autocorrelations
- fix up "delete" command ("insufficient args" bug)
- clean up gnuplot background code for gnuplot controller
- delete extra variables after chow test
- minor fixes to LaTeX printing
- use proper pointers to keep track of multiple windows
- add "max p-value" statement to regression printout
- put back Durbin-Watson printout line for hccm, tsls
- fix inconsistency with docs in setting of the break point
for the Chow test
- more fixes to csv file reading
- bug fix in freq plot after freq print
- update sample label when excluding missing obs
- update iconfilesel and gtksheet to gtk+extra-0.99.11
- configure: use installed gtkextra if available
- substantial code cleanups and speedups in GUI.
06/14/2000 version 0.2i
- improve consistency in GUI w.r.t. opening of data files
and databases
- update manual and help files
- add "delete" command (trash last variable)
- fix potential data files paths bug
- fix hsk bug (aux regression could fail silently)
- add some data files from William Greene's book
- add "while" syntax to command loop mechanism
(useful for iterated least squares -- tested on Greene's
example 11.3 of nonlinear estimation). See manual.
- fix sample-length bug in add, omit
- some more informative error messages from genr
- activate "add" and "omit" for logit, probit
- make covariance matrix available for logit, probit
- add more error trapping for "add"
- code cleanup in compare.c
- can capture log-likelihood in genr with "$lnl" after logit, probit
06/09/2000 version 0.2h
- add estimation of logit and probit models and update manual
- fixes to "configure" script (gtk, termcap, paths.h) after
compilation on HPUX
- eslclient creates user dir if it doesn't exist
- add (experimental) "parenthesizer" for genr expressions
- rearrange "summary" output
- rename Windows dll
- fix bug in "fcast" command
06/01/2000 version 0.2g
- "smpl -o" drops all observations with missing values
- binary datafiles can be saved in single or double precision
(-s or -o flags)
- Binary datafile saving added to GUI
- Can add case markers from file to an open data set
- allow for a comment line at the top of a database .idx file
(and display it in GUI)
- code cleanups
- delete redundant stuff
- improve some string handling functions
- fix win32 path-searching bug?
05/24/2000 version 0.2f
- updates to manual
- add handler for Penn World Table
- improve handling of native databases
- reformat some output
- add % operator in genr
- add -o flag to "smpl" command for sampling via a dummy
variable mask.
- GUI: where appropriate, clicking on a var in the main window
will pop its ID # into an open dialog box.
- add fcasterr to GUI, and improve graphing of forecasts
- move setobs from Data to Sample menu
- add "Gretl console" to GUI (under File menu) and update manual
- add to path-searching for datafiles called from scripts
(include the script dir in the search)
05/10/2000 version 0.2e
- minor changes to GUI menus
- add xpm icon for X
- correct printout of t-stats for AR coefficients
- correct wasteful malloc in xpxxpy function
- better printout for F-statistic
- add discussion of NIST reference datasets to manual
- add NIST datasets to package
- add some error-checking and reporting for datafile opening
- various fixes to "genr"
- search for libesl.cfg in dir where executable resides
(win32, Dirk's idea)
- improve handling of constants in corr, corrgm
- fix error messages from corc_hilu
- rename eslclnt.exe as eslclient.exe (win32)
- fix paths in win32 (helpfile and executable, for rerunning
commands)
05/05/2000 version 0.2d
- improvements to manual
- fix wrong error message for zero weight var in wls
- explicit check for degeneracy in figuring summary stats
- add median(), var(), cov() to functions in genr
- add "spearman" command (rank correlation)
- add "runs" command (runs test for randomness)
- fix bug in calculation of rho hat
- fix lengths of series created by diff, etc, in relation to smpl
05/01/2000 version 0.2c
- add setobs command
- add seed command
- manual: add section on panel data, improve "genr" section
- more CSV bugfixes
- make adf, coint commands more informative
- add "paneldum" (for panel data) to genr command
04/28/2000 version 0.2b
- add fcasterr command
- many fixes in response to comments from Ramu Ramanathan and
Dirk Eddelbuettel
- improve Makefiles
- bugfix in binary datafile handling
- update manual
04/22/2000 version 0.2a
- add arch, sim and multiply commands
- many bugfixes after running numerous test files
- further code cleanups
- improvements in printing forecasts
- fix memory leaks
04/14/2000 version 0.2
- many fixes for greater consistency with esl
- printout improvements
- internal code cleanups
- bug fixes (notably for CSV file handling)
04/05/2000 version 0.1z
- add a gui controller for gnuplot graphs
- many bugfixes
- reorganize (simplify) command-line arguments
- more updates to documentation
03/27/2000 version 0.1y
- improvements and bugfixes to CSV file handling
- improvements to "add" and "omit" commands
- add "coint" command (cointegration test)
03/23/2000 version 0.1x
- many small bugfixes
- add GNU autoconf apparatus for configuring/compiling
- updates to documentation
02/29/2000 version 0.1w
- fixes to build process
- add iconic file selector
- add "export CSV" function
- bug fixes to database reading routines
02/22/2000 version 0.1v
- add White's standard errors (hccm command)
- update help files
- Make model stats (e.g. ess) available for adding to data
set in gui, under Model data menu.
- Add option to display values of selected variables, as well as "all".
- Fix pvalue in command-line client so that it does not act
interactively when supplied with all params on the command line.
- Add facility to edit the full data header (.hdr) file in
gretl (under Data menu).
- Activate the "Add observation" feature in ssheet; add a
similar "Insert Observation" option. Minimal testing so far.
- Fix (I think) bug in setup of new spreadsheet data set.
- Give "infobox" feedback when importing database series.
02/13/2000 version 0.1u
- save data files in GNU Octave format
- updates to documentation
02/08/2000 version 0.1t
- gretl adopted as GNU program
- add context-sensitive help for command dialogs
- add file, save for model output windows
- add search facility for help file, database windows
02/02/2000 version 0.1s
- Fixes to Makefiles
- Improvements to loop mode (Monte Carlo simulations)
01/31/2000 version 0.1r
- Initial public release
|