Function packages, written in gretl's scripting language hansl, extend the program's functionality in numerous econometric domains.
There are two types of packages:
Addons: these official
packages are written by members
of the gretl team. While gretl offers no warranty, the addons are
tested as thoroughly as the core gretl code. They are supplied in
the installers we provide for MS Windows and macOS, and also in the
gretl packages for Linux prepared
by Debian
and Fedora.
The addons are automatically present in the gretl GUI. For example,
the menu item Model/Univariate time series/GARCH variants
gives an
interface to the gig
(Garch in gretl) addon.
Contributed packages: these are written by gretl users who may
or may not be team members. They are moderated by the team, to ensure
that they work
(that is, produce plausible output when called
as per their documentation), that they are reasonably well documented,
and that they are non-redundant (don't just replicate functionality
that's already available). As of this writing over 170 such packages
are available. These packages are not installed automatically with the
rest of gretl. You can install them from the GUI program: go to the
menu item File/Function
packages/On server
to see a listing of what's available.
Some such packages can only be used via the command line, but many of them
also include a GUI hook
(that is, offer to insert a menu item
from which they can be called via a dialog box).
Assorted statistical methods: correspondence analysis, multidimensional scaling, factor analysis, CUB models, weighted statistics; additional tests (Bartlett, Friedman, Kruskal-Wallis, Kolmogorov-Smironv, Levene).
Bayesian methods: Bayesian regression, Bayesian VARs, Bayesian Model Averaging.
Causal inference: matching estimators.
Additional tools for univariate models: multipliers for ARDL models, time-varying parameters, spatial regression, threshold models, Bai-Perron structural breaks, marginal effects for nonlinear models, Box-Cox modelling, extra utilities for IV estimation.
Microeconometric methods: probit models with heteroskedasticity and IV, ROC curve and specification tests for binary models, discrete choice models, predictions for Tobit, ordered, count data models.
Machine learning tools: Regularized least squares (LASSO, RIDGE, Elastic Net), adaptive and graphical LASSO, k-means clustering, random forests.
Time series:
Univariate: time series filters (Christiano-Fitzgerald, Hamilton), Harvey-style structural models, additional unit root tests (including Phillips-Perron and HEGY), FM-OLS, SETAR models.
Multivariate: Structural VARs, recursive VARs, Dynamic Factor Models (also hierarchical), multivariate spectra, Markov-switching VARs, bootstrap-based cointegration tests.
Financial: Automated data downloading from Yahoo Finance, univariate GARCH and SV models, DCC and BEKK multivariate models.
Forecasts: Forecats evaluations, Diebold-Mariano test, interval forecasts.
Models for panel data: IV models, dynamic binary models, FE Poisson, Pooled and Mean Group estimator for dynamic models, RE Tobit, cross-sectional tests, unit-root and cointegration tests.
Specialized plotting methods: geographical maps, candlesticks, heamaps, scatterplots with factor separation, bivariate and factorized densities.
Various utilities for Hansl programming.
Entries in boldface are provided via addons, and are therefore immediately available upon successful installation on most platforms.
You can browse automatically updated listings of the gretl addons here and the contributed packages here. In both cases you can access documentation for the packages.
Gretl users who are familiar with hansl are welcome to contribute packages. There's a comprehensive document explaining how to do this and also, thanks to the late Stefano Fachin, a nice quick guide.
Allin Cottrell and Riccardo (Jack) Lucchetti