genfuncs

## Functions

 int sort_series () int gretl_sort_by () int rank_series () gretl_matrix * rank_vector () int diff_series () int orthdev_series () int cum_series () int resample_series () int block_resample_series () int fracdiff_series () int boxcox_series () int filter_series () gretl_matrix * filter_matrix () int exponential_movavg_series () int movavg_series () int seasonally_adjust_series () int panel_statistic () gretl_matrix * panel_shrink () int hp_filter () int bkbp_filter () int butterworth_filter () int poly_trend () int weighted_poly_trend () void poly_weights () gretl_matrix * hp_gain () gretl_matrix * butterworth_gain () int dummy () int panel_dummies () int gen_unit () int panel_unit_first_obs () int gen_time () int gen_wkday () const double * gretl_plotx () double * get_fit_or_resid () int get_observation_number () int get_t_from_obs_string () int list_linear_combo () double imhof () double dw_pval () gretl_matrix * multi_acf () gretl_matrix * multi_xcf () gretl_matrix * forecast_stats () double gretl_round () double gretl_bessel () double gretl_npv () double gretl_irr () double logistic_cdf () gretl_matrix * matrix_chowlin () int list_ok_dollar_vars () int nadaraya_watson () int gretl_loess () double series_get_nobs () double series_sum_all () gretl_matrix * aggregate_by () int fill_dataset_dates_series () int fill_day_of_week_array ()

## Functions

### sort_series ()

int
sort_series (const double *x,
double *y,
int f,
const DATASET *dset);

### gretl_sort_by ()

int
gretl_sort_by (const double *x,
const double *y,
double *z,
const DATASET *dset);

### rank_series ()

int
rank_series (const double *x,
double *y,
int f,
const DATASET *dset);

### rank_vector ()

gretl_matrix *
rank_vector (const gretl_matrix *x,
int f,
int *err);

### diff_series ()

int
diff_series (const double *x,
double *y,
int f,
const DATASET *dset);

Calculates the differenced counterpart to the input series x . If f = F_SDIFF, the seasonal difference is computed; if f = F_LDIFF, the log difference, and if f = F_DIFF, the ordinary first difference.

#### Parameters

 x array of original data. y array into which to write the result. f function, F_DIFF, F_SDIFF or F_LDIFF. dset data set information.

#### Returns

0 on success, non-zero error code on failure.

### orthdev_series ()

int
orthdev_series (const double *x,
double *y,
const DATASET *dset);

Calculates in y the forward orthogonal deviations of the input series x . That is, y[t] is the scaled difference between x[t] and the mean of the subsequent observations on x.

#### Parameters

 x array of original data. y array into which to write the result. dset data set information.

#### Returns

0 on success, non-zero error code on failure.

### cum_series ()

int
cum_series (const double *x,
double *y,
const DATASET *dset);

### resample_series ()

int
resample_series (const double *x,
double *y,
const DATASET *dset);

### block_resample_series ()

int
block_resample_series (const double *x,
double *y,
int blocklen,
const DATASET *dset);

### fracdiff_series ()

int
fracdiff_series (const double *x,
double *y,
double d,
int diff,
int obs,
const DATASET *dset);

Calculates the fractionally differenced or lagged counterpart to the input series x . The fractional difference operator is defined as (1-L)^d, while the fractional lag operator 1-(1-L)^d.

#### Parameters

 x array of original data. y array into which to write the result. d fraction by which to difference. diff boolean variable 1 for fracdiff, 0 for fraclag obs used for autoreg calculation, -1 if whole series should be calculated otherwise just the observation for obs is calculated dset data set information.

#### Returns

0 on success, non-zero error code on failure.

### boxcox_series ()

int
boxcox_series (const double *x,
double *y,
double d,
const DATASET *dset);

Calculates in y the Box-Cox transformation for the input series x .

#### Parameters

 x array of original data. y array into which to write the result. d lambda parameter. dset data set information.

#### Returns

0 on success, non-zero error code on failure.

### filter_series ()

int
filter_series (const double *x,
double *y,
const DATASET *dset,
gretl_matrix *A,
gretl_matrix *C,
double y0);

Filters x according to y_t = C(L)/A(L) x_t. If the intended AR order is p, A should be a vector of length p. If the intended MA order is q, C should be vector of length (q+1), the first entry giving the coefficient at lag 0. However, if C is NULL this is taken to mean that the lag-0 MA coefficient is unity (and all others are zero).

#### Parameters

 x array of original data. y array into which to write the result. dset data set information. A vector for autoregressive polynomial. C vector for moving average polynomial. y0 initial value of output series.

#### Returns

0 on success, non-zero error code on failure.

### filter_matrix ()

gretl_matrix *
filter_matrix (gretl_matrix *X,
gretl_vector *A,
gretl_vector *C,
double y0,
int *err);

Filters the columns of x according to y_t = C(L)/A(L) x_t. If the intended AR order is p, A should be a vector of length p. If the intended MA order is q, C should be vector of length (q+1), the first entry giving the coefficient at lag 0. However, if C is NULL this is taken to mean that the lag-0 MA coefficient is unity (and all others are zero).

#### Parameters

 X matrix of original data. A vector for autoregressive polynomial. C vector for moving average polynomial. y0 initial value of output series.

#### Returns

0 on success, non-zero error code on failure.

### exponential_movavg_series ()

int
exponential_movavg_series (const double *x,
double *y,
const DATASET *dset,
double d,
int n);

#### Parameters

 x array of original data. y array into which to write the result. dset data set information. d coefficient on lagged x . n number of x observations to average to give the initial y value.

#### Returns

0 on success, non-zero error code on failure.

### movavg_series ()

int
movavg_series (const double *x,
double *y,
const DATASET *dset,
int k,
int center);

#### Parameters

 x array of original data. y array into which to write the result. dset data set information. k number of terms in MA. center if non-zero, produce centered MA.

#### Returns

0 on success, non-zero error code on failure.

int
seasonally_adjust_series (const double *x,
double *y,
DATASET *dset,
int tramo,
int use_log);

### panel_statistic ()

int
panel_statistic (const double *x,
double *y,
const DATASET *dset,
int k,
const double *mask);

Given the data in x , constructs in y a series containing a panel-data statistic.

#### Parameters

 x source data. y target into which to write. dset data set information. k code representing the desired statistic: F_PNOBS, F_PMIN, F_PMAX, F_PSUM, F_PMEAN, F_PXSUM or F_PSD. mask either NULL or a series with 0s for observations to be excluded from the calculations, non-zero values at other observations.

#### Returns

0 on success, non-zero on error.

### panel_shrink ()

gretl_matrix *
panel_shrink (const double *x,
const DATASET *dset,
int *err);

Constructs a column vector holding the first non-missing observation of x for each panel unit within the current sample range. If a unit has no valid observations it is skipped.

#### Parameters

 x panel-data source series. dset data set information. err location to receive error code.

#### Returns

a new column vector, or NULL on error.

### hp_filter ()

int
hp_filter (const double *x,
double *hp,
const DATASET *dset,
double lambda,
gretlopt opt);

Calculates the "cycle" component of the time series in array x , using the Hodrick-Prescott filter. Adapted from the original FORTRAN code by E. Prescott.

#### Parameters

 x array of original data. hp array in which filtered series is computed. dset data set information. lambda smoothing parameter (or NADBL to use the default value). opt if OPT_T, return the trend rather than the cycle.

#### Returns

0 on success, non-zero error code on failure.

### bkbp_filter ()

int
bkbp_filter (const double *x,
double *bk,
const DATASET *dset,
int bkl,
int bku,
int k);

Calculates the Baxter and King bandpass filter.

#### Parameters

 x array of original data. bk array into which to write the filtered series. dset data set information. bkl lower frequency bound (or 0 for automatic). bku upper frequency bound (or 0 for automatic). k approximation order (or 0 for automatic).

#### Returns

0 on success, non-zero error code on failure.

### butterworth_filter ()

int
butterworth_filter (const double *x,
double *bw,
const DATASET *dset,
int n,
double cutoff);

Calculates the Butterworth filter. The code that does this is based on D.S.G. Pollock's IDEOLOG -- see http://www.le.ac.uk/users/dsgp1/

#### Parameters

 x array of original data. bw array into which to write the filtered series. dset data set information. n desired lag order. cutoff desired angular cutoff in degrees (0, 180).

#### Returns

0 on success, non-zero error code on failure.

### poly_trend ()

int
poly_trend (const double *x,
double *fx,
const DATASET *dset,
int order);

Calculates a trend via the method of orthogonal polynomials. Based on C code for D.S.G. Pollock's DETREND program.

#### Parameters

 x array of original data. fx array into which to write the fitted series. dset data set information. order desired polynomial order.

#### Returns

0 on success, non-zero error code on failure.

### weighted_poly_trend ()

int
weighted_poly_trend (const double *x,
double *fx,
const DATASET *dset,
int order,
gretlopt opt,
double wratio,
double midfrac);

Calculates a trend via the method of orthogonal polynomials, using the specified weighting scheme. Based on C code for D.S.G. Pollock's DETREND program.

#### Parameters

 x array of original data. fx array into which to write the fitted series. dset data set information. order desired polynomial order. opt weighting option (OPT_Q = quadratic, OPT_B = cosine bell, OPT_C = crenelated). wratio ratio of maximum to minimum weight. midfrac proportion of the data to be treated specially, in the middle.

#### Returns

0 on success, non-zero error code on failure.

### poly_weights ()

void
poly_weights (double *w,
int T,
double wmax,
double midfrac,
gretlopt opt);

Calculates a set of weights; intended for use with polynomial trend fitting.

#### Parameters

 w array into which the weights will be written. T the length of w . wmax the ratio of maximum to minimum weight. midfrac the size of the central section that should be given the minimum weight. opt weighting scheme option (OPT_Q = quadratic, OPT_B = cosine bell, OPT_C = crenelated).

### hp_gain ()

gretl_matrix *
hp_gain (double lambda,
int hipass);

#### Parameters

 lambda H-P parameter. hipass 1 for high-pass filter, 0 for low-pass.

#### Returns

a matrix holding omega values from 0 to \pi in column 0, and the corresponding filter gain in column 1.

### butterworth_gain ()

gretl_matrix *
butterworth_gain (int n,
double cutoff,
int hipass);

#### Parameters

 n order of the filter. cutoff angular cutoff in radians. hipass 1 for high-pass filter, 0 for low-pass.

#### Returns

a matrix holding omega values from 0 to \pi in column 0, and the corresponding filter gain in column 1.

### dummy ()

int
dummy (DATASET *dset,
int center);

Adds to the data set (if these variables are not already present) a set of periodic (usually seasonal) dummy variables.

#### Parameters

 dset dataset struct. center if greater than zero subtract the population mean from each of the generated dummies; if less than zero, do not subtract the mean but generate dummies with labels on the same pattern as centered dummies (for internal use in VECMs). Usually this argument is set to zero.

#### Returns

the ID number of the first dummy variable on success, or 0 on error.

### panel_dummies ()

int
panel_dummies (DATASET *dset,
gretlopt opt,
PRN *prn);

Adds to the data set a set of dummy variables corresponding to either the cross-sectional units in a panel, or the time periods.

#### Parameters

 dset dataset struct. opt OPT_T for time dummies, otherwise unit dummies. prn printer for warning, or NULL.

#### Returns

0 on successful completion, error code on error.

### gen_unit ()

int
gen_unit (DATASET *dset);

(For panel data only) adds to the data set an index variable that uniquely identifies the cross-sectional units.

#### Parameters

 dset dataset struct.

#### Returns

0 on successful completion, error code on error.

### panel_unit_first_obs ()

int
panel_unit_first_obs (int t,
const DATASET *dset);

#### Parameters

 t zero-based observation number. dset data information struct.

#### Returns

1 if observation t is the first time-series observation on a given cross-sectional unit in a panel dataset, else 0.

### gen_time ()

int
gen_time (DATASET *dset,
int tm);

Generates (and adds to the dataset, if it's not already present) a time-trend or index variable. This function is panel-data aware: if the dataset is a panel and tm is non-zero, the trend will not simply run consecutively over the entire range of the data, but will correctly represent the location in time of each observation. The index is 1-based.

#### Parameters

 dset dataset struct. tm if non-zero, an actual time trend is wanted, otherwise just an index of observations.

#### Returns

0 on success, non-zero on error.

### gen_wkday ()

int
gen_wkday (DATASET *dset);

### gretl_plotx ()

const double *
gretl_plotx (const DATASET *dset,
gretlopt opt);

Finds or creates a special dummy variable for use on the x-axis in plotting; this will have the full length of the data series as given in dset , and will be appropriately configured for the data frequency. Do not try to free this variable.

#### Parameters

 dset data information struct. opt can include OPT_P for panel time-series plot.

#### Returns

pointer to plot x-variable, or NULL on failure.

### get_fit_or_resid ()

double *
get_fit_or_resid (const MODEL *pmod,
DATASET *dset,
ModelDataIndex idx,
char *vname,
char *vlabel,
int *err);

Creates a full-length array holding the specified model data, and writes name and description into the vname and vlabel .

#### Parameters

 pmod pointer to source model. dset information on the data set. idx vname location to write series name (length VNAMELEN) vlabel location to write series description (length should be MAXLABEL). err location to receive error code.

#### Returns

allocated array on success or NULL on failure.

### get_observation_number ()

int
get_observation_number (const char *s,
const DATASET *dset);

### get_t_from_obs_string ()

int
get_t_from_obs_string (const char *s,
const DATASET *dset);

### list_linear_combo ()

int
list_linear_combo (double *y,
const int *list,
const gretl_vector *b,
const DATASET *dset);

### imhof ()

double
imhof (const gretl_matrix *m,
double arg,
int *err);

### dw_pval ()

double
dw_pval (const gretl_matrix *u,
const gretl_matrix *X,
double *pDW,
int *err);

### multi_acf ()

gretl_matrix *
multi_acf (const gretl_matrix *m,
const int *list,
const DATASET *dset,
int p,
int *err);

### multi_xcf ()

gretl_matrix *
multi_xcf (const void *px,
int xtype,
const void *py,
int ytype,
const DATASET *dset,
int p,
int *err);

### forecast_stats ()

gretl_matrix *
forecast_stats (const double *y,
const double *f,
int t1,
int t2,
gretlopt opt,
int *err);

### gretl_round ()

double
gretl_round (double x);

### gretl_bessel ()

double
gretl_bessel (char type,
double v,
double x,
int *err);

### gretl_npv ()

double
gretl_npv (int t1,
int t2,
const double *x,
double r,
int pd,
int *err);

### gretl_irr ()

double
gretl_irr (const double *x,
int n,
int pd,
int *err);

### logistic_cdf ()

double
logistic_cdf (double x);

### matrix_chowlin ()

gretl_matrix *
matrix_chowlin (const gretl_matrix *Y,
const gretl_matrix *X,
int f,
int *err);

Interpolate, from annual to quarterly or quarterly to monthly, via the Chow-Lin method. See Gregory C. Chow and An-loh Lin, "Best Linear Unbiased Interpolation, Distribution, and Extrapolation of Time Series by Related Series", The Review of Economics and Statistics, Vol. 53, No. 4 (November 1971) pp. 372-375.

If X is provided, it must have T * f rows.

#### Parameters

 Y T x k: holds the original data to be expanded, series in columns. X (optionally) holds covariates of Y at the higher frequency: if these are supplied they supplement the default set of regressors, namely, constant plus quadratic trend. f the expansion factor: 3 for quarterly to monthly or 4 for annual to quarterly. Only these factors are supported. err location to receive error code.

#### Returns

matrix containing the expanded series, or NULL on failure.

### list_ok_dollar_vars ()

int
list_ok_dollar_vars (DATASET *dset,
PRN *prn);

#### Returns

int
nadaraya_watson (const double *y,
const double *x,
double h,
DATASET *dset,
double *m);

Implements the Nadaraya-Watson non-parametric estimator for the conditional mean of y given x via the formula

\widehat{m}_h(x)=\frac{\sum_{i=1}^n K_h(x-X_i) Y_i}{\sum_{i=1}^nK_h(x-X_i)}

and computes it for all elements of x . Note that, in principle, the kernel K(X_i-X_j) must be computed for every combination of i and j, but since the function K() is assumed to be symmetric, we compute it once to save time.

The scalar h holds the kernel bandwidth; if negative, it implies that the leave-one-out estimator (essentially a jackknife estimator; see Pagan and Ullah, Nonparametric Econometrics, page 119) is wanted. A rudimentary form of trimming is implemented, but it will have to be refined.

#### Parameters

 y array with "dependent variable" x array with "explanatory variable" h double, bandwidth (may be negative; see below) dset data set information. m array to hold results

#### Returns

0 on successful completion, non-zero code on error.

### gretl_loess ()

int
gretl_loess (const double *y,
const double *x,
int poly_order,
double bandwidth,
gretlopt opt,
DATASET *dset,
double *m);

### series_get_nobs ()

double
series_get_nobs (int t1,
int t2,
const double *x);

### series_sum_all ()

double
series_sum_all (int t1,
int t2,
const double *x);

### aggregate_by ()

gretl_matrix *
aggregate_by (const double *x,
const double *y,
const int *xlist,
const int *ylist,
const char *fncall,
const DATASET *dset,
int *err);

Aggregates one or more data series (x) "by" the values of one or more discrete series (y). In general either x or xlist should be non-NULL, and one of y or ylist should be non-NULL. (If xlist is non-NULL then x is ignored, and similarly for ylist and y ). For an account of the matrix that is returned, see the help for gretl's "aggregate" command.

#### Parameters

 x data array. y discrete variable. xlist list of x series or NULL. ylist list of y series or NULL. fncall the name of the aggregation function. dset data set information. err location to receive error code.

#### Returns

allocated matrix, or NULL on failure.

### fill_dataset_dates_series ()

int
fill_dataset_dates_series (const DATASET *dset,
double *x);

### fill_day_of_week_array ()

int
fill_day_of_week_array (double *dow,
const double *y,
const double *m,
const double *d,
const DATASET *dset);