Limdep: Libgretl Reference Manual

Limdep

Limdep — models for limited dependent variables and related cases.

Functions

MODEL	binary_logit ()
MODEL	binary_probit ()
MODEL	ordered_logit ()
MODEL	ordered_probit ()
MODEL	multinomial_logit ()
MODEL	biprobit_model ()
MODEL	reprobit_model ()
MODEL	logistic_model ()
MODEL	interval_model ()
MODEL	tobit_model ()
MODEL	duration_model ()
MODEL	count_model ()
MODEL	heckit_model ()
int	fishers_exact_test ()
double	ordered_model_prediction ()
int	logistic_ymax_lmax ()
gretl_matrix *	mn_logit_probabilities ()
gretl_matrix *	ordered_probabilities ()
double	mn_logit_prediction ()
void	binary_model_hatvars ()

Includes

#include <libgretl.h>

Description

Covers logit (binary, ordered or multinomial), probit (binary or ordered), logistic, tobit, interval regression, models for count data and for duration data, and the heckit sample-selection model. Plus a few utility functions.

Functions

binary_logit ()

MODEL
binary_logit (const int *list,
              DATASET *dset,
              gretlopt opt,
              PRN *prn);

Computes estimates of the logit model specified by list , using Newton-Raphson.

Parameters

list	binary dependent variable plus list of regressors.
dset	dataset struct.
opt	if includes OPT_R form robust (QML) estimates of standard errors and covariance matrix; if OPT_P arrange for printing of p-values, not slopes at mean; if OPT_A treat as an auxiliary regression.
prn	printing struct in case additional information is wanted (in which case add OPT_V to `opt` ).

Returns

a MODEL struct, containing the estimates.

binary_probit ()

MODEL
binary_probit (const int *list,
               DATASET *dset,
               gretlopt opt,
               PRN *prn);

Computes estimates of the probit model specified by list , using Newton-Raphson.

Parameters

list	binary dependent variable plus list of regressors.
dset	dataset struct.
opt	if includes OPT_R form robust (QML) estimates of standard errors and covariance matrix; if OPT_P arrange for printing of p-values, not slopes at mean; if OPT_A treat as an auxiliary regression.
prn	printing struct in case additional information is wanted (in which case add OPT_V to `opt` ).

Returns

a MODEL struct, containing the estimates.

ordered_logit ()

MODEL
ordered_logit (const int *list,
               DATASET *dset,
               gretlopt opt,
               PRN *prn);

Computes ML estimates of the ordered logit model specified by list .

Parameters

list	ordinal dependent variable plus list of regressors.
dset	dataset struct.
opt	if includes OPT_R form robust (QML) estimates of standard errors and covariance matrix; if includes OPT_V, produce verbose output.
prn	printing struct in case additional information is wanted (OPT_V).

Returns

a MODEL struct, containing the estimates.

ordered_probit ()

MODEL
ordered_probit (const int *list,
                DATASET *dset,
                gretlopt opt,
                PRN *prn);

Computes ML estimates of the ordered probit model specified by list .

Parameters

list	ordinal dependent variable plus list of regressors.
dset	dataset struct.
opt	if includes OPT_R form robust (QML) estimates of standard errors and covariance matrix; if includes OPT_V, produce verbose output.
prn	printing struct in case additional information is wanted (OPT_V).

Returns

a MODEL struct, containing the estimates.

multinomial_logit ()

MODEL
multinomial_logit (const int *list,
                   DATASET *dset,
                   gretlopt opt,
                   PRN *prn);

Computes ML estimates of the multinomial model specified by list .

Parameters

list	discrete dependent variable plus list of regressors.
dset	dataset struct.
opt	if includes OPT_R form robust (QML) estimates of standard errors and covariance matrix; if includes OPT_V, produce verbose output.
prn	printing struct in case additional information is wanted (OPT_V).

Returns

a MODEL struct, containing the estimates.

biprobit_model ()

MODEL
biprobit_model (const int *list,
                DATASET *dset,
                gretlopt opt,
                PRN *prn);

Computes estimates of the bivariate probit model specified by list , using maximum likelihood via Newton-Raphson.

Parameters

list	binary dependent variable 1, binary dependent variable 2, list of regressors for y1. If `list` ends here, it is assumed that the explanatory variables for y2 are the same as y1. Otherwise, the list must include a separator and the list of regressors for y2.
dset	dataset struct.
opt	can contain OPT_Q for quiet operation, OPT_V for verbose operation, OPT_R for robust covariance matrix, OPT_G for covariance matrix based on Outer Product of Gradient.
prn	printing struct.

Returns

a MODEL struct, containing the estimates.

reprobit_model ()

MODEL
reprobit_model (const int *list,
                DATASET *dset,
                gretlopt opt,
                PRN *prn);

Produce random-effects probit estimates of the model given in list .

Parameters

list	dependent variable plus list of regressors.
dset	dataset struct.
opt	option flags (may include OPT_V for verbose output).
prn	printing struct for iteration info (or NULL if this is not wanted).

Returns

a MODEL struct, containing the estimates.

logistic_model ()

MODEL
logistic_model (const int *list,
                double lmax,
                DATASET *dset,
                gretlopt opt);

Estimate the model given in list using the logistic transformation of the dependent variable.

Parameters

list	dependent variable plus list of regressors.
lmax	value for the asymptote of the logistic curve, or `NADBL` for automatic treatment of this.
dset	pointer to dataset.
opt	option flags.

Returns

a MODEL struct, containing the estimates.

interval_model ()

MODEL
interval_model (const int *list,
                DATASET *dset,
                gretlopt opt,
                PRN *prn);

Parameters

list	high/low (2 variables) plus list of regressors.
dset	dataset struct.
opt	if includes `OPT_R` form robust (QML) estimates of standard errors and covariance matrix; if includes `OPT_V` give verbose operation.
prn	printing struct in case additional information is wanted (signalled via `OPT_V`).

Returns

a MODEL struct, containing interval estimates of the model specified by list .

tobit_model ()

MODEL
tobit_model (const int *list,
             double llim,
             double rlim,
             DATASET *dset,
             gretlopt opt,
             PRN *prn);

Produce Tobit estimates of the model given in list .

Parameters

list	dependent variable plus list of regressors.
llim	left bound on dependent variable; use NADBL for no left-censoring.
rlim	right bound on dependent variable; use NADBL for no right-censoring.
dset	dataset struct.
opt	may include OPT_V for verbose operation, OPT_R for robust (QML) standard errors.
prn	printing struct for iteration info (or NULL if this is not wanted).

Returns

a MODEL struct, containing the estimates.

duration_model ()

MODEL
duration_model (const int *list,
                DATASET *dset,
                gretlopt opt,
                PRN *prn);

Estimate the duration model given in list using ML.

Parameters

list	dependent variable plus list of regressors.
dset	dataset struct.
opt	may include OPT_R for robust covariance matrix.
prn	printing struct for iteration info (or NULL is this is not wanted).

Returns

a MODEL struct, containing the estimates.

count_model ()

MODEL
count_model (const int *list,
             int ci,
             DATASET *dset,
             gretlopt opt,
             PRN *prn);

Estimate the count data model given in list using ML.

Parameters

list	dependent variable plus list of regressors.
ci	either POISSON or NEGBIN.
dset	dataset struct.
opt	may include OPT_R for robust covariance matrix.
prn	printing struct for iteration info (or NULL is this is not wanted).

Returns

a MODEL struct, containing the estimates.

heckit_model ()

MODEL
heckit_model (const int *list,
              DATASET *dset,
              gretlopt opt,
              PRN *prn);

Produce Heckit estimates of the model given in list . The list must include a separator to divide the main equation from the selection equation.

Parameters

list	dependent variable plus list of regressors.
dset	dataset struct.
opt	option flags (may include OPT_V for verbose output).
prn	printing struct for iteration info (or NULL is this is not wanted).

Returns

a MODEL struct, containing the estimates.

fishers_exact_test ()

int
fishers_exact_test (const Xtab *tab,
                    PRN *prn);

Computes and prints to prn the p-value for Fisher's Exact Test for association between the two variables represented in tab .

Parameters

tab	pointer to 2 x 2 cross-tabulation struct.
prn	gretl printer.

Returns

0 on successful completion, error code on error.

ordered_model_prediction ()

double
ordered_model_prediction (const MODEL *pmod,
                          double Xb,
                          int ymin);

Parameters

pmod	model for ordered data, either logit or probit.
Xb	X\beta, the value of the index function at a given observation.
ymin	the minimum value of the dependent variable.

Returns

the "predicted value" of the (ordinal) dependent variable, taken to be the value for which the estimated probability is greatest.

logistic_ymax_lmax ()

int
logistic_ymax_lmax (const double *y,
                    const DATASET *dset,
                    double *ymax,
                    double *lmax);

Checks that the non-missing values of y are all positive, and if so writes the maximum value of y to ymax . The value written to lmax is 1 if max(y) < 1, else 100 if max(y) < 100, else 1.1 * max(y).

Parameters

y	data series.
dset	dataset information.
ymax	location to receive max(y).
lmax	location to receive a guess at a suitable asymptote for a logistic curve fitted to `y` .

Returns

0 on success, non-zero on error.

mn_logit_probabilities ()

gretl_matrix *
mn_logit_probabilities (const MODEL *pmod,
                        int t1,
                        int t2,
                        const DATASET *dset,
                        int *err);

Computes the estimated probabilities of the outcomes for a multinomial logit model. The returned matrix is n x m, where n is the number of observations in the sample range over which the model was estimated and m is the number of distinct outcomes. Each element represents the conditional probability of outcome j given the values of the regressors at observation i.

If any of the regressor values are missing at a given observation the probability is set to NaN; provided the regressor information is complete we compute the outcome probabilities even if the actual outcome is missing.

Parameters

pmod	pointer to multinomial logit model
dset	dataset struct.
err	location to receive error code.

Returns

allocated matrix or NULL on failure.

ordered_probabilities ()

gretl_matrix *
ordered_probabilities (const MODEL *pmod,
                       const double *zhat,
                       int t1,
                       int t2,
                       const DATASET *dset,
                       int *err);

mn_logit_prediction ()

double
mn_logit_prediction (const gretl_matrix *Xt,
                     const double *b,
                     const gretl_matrix *yvals);

Parameters

Xt	vector of regressors at observation t.
b	array of coefficients.
yvals	vector of dependent variable values.

Returns

the predicted value of the dependent variable, that is, the value for which the estimated probability is greatest.

binary_model_hatvars ()

void
binary_model_hatvars (MODEL *pmod,
                      const gretl_matrix *ndx,
                      const int *y,
                      gretlopt opt);