gretl is an application that has been designed for econometric analysis. It is written in the C programming language and is also available for the MacOS X platform.
Key features include:
- Intuitive user interface.
- Multilingual support.
- A wide variety of estimators including: least squares, maximum likelihood, GMM; single-equation and system methods.
- Time series methods: ARIMA, a wide variety of univariate GARCH-type models, VARs and VECMs (including structural VARs), unit-root and cointegration tests, Kalman filter, etc.
- Limited dependent variables: logit, probit, tobit, sample selection, interval regression, models for count and duration data, etc.
- Panel-data estimators, including instrumental variables, probit and GMM-based dynamic panel models.
- Output models as LaTeX files, in tabular or equation format.
- Integrated powerful scripting language (known as hansl), with a wide range of programming tools and matrix operations.
- GUI controller for fine-tuning Gnuplot graphs.
- An expanding range of contributed function packages, written in hansl.
- Facilities for easy exchange of data and results with GNU R, GNU Octave, Python, Ox and Stata.
Supported formats include:
- Own XML data files.
- Comma Separated Values.
- Gnumeric and Open Document worksheets.
- Stata .dta files.
- SPSS .sav files.
- Eviews workfiles.
- JMulTi data files.
- Own format binary databases (allowing mixed data frequencies and series lengths).
- RATS 4 databases and PC-Give databases.
gretl has language support for English, French, Italian, Spanish, Polish, German, Basque, Catalan, Galician, Portuguese, Russian, Turkish, Czech, Traditional Chinese, Albanian, Bulgarian, Greek, Japanese and Romanian.
New accessor, $tmax: gives the maximum legal setting for the end of the sample range via "smpl".
$model bundle: add members t1 and t2 (start and end of estimation range).
New function getinfo(): returns a bundle with information on a specified series.
New function cdummify(): applies to a list, with the effect of "dummifying" any and all series in the list that have the "coded" attribute set.
New function numhess() to compute numerical Hessian; also improve accuracy of the underlying numerical_hessian code.
"pca" command and princomp() function: center but do not standardize when computing principal components using the covariance matrix.
"midasreg" command, mds() term: allow specification of min and max lags via scalar variables.
psdroot() function: use improved algorithm (from Golub and Van Loan).
fdjac() function: add (optional) trailing argument to set step size.
randgen() function (and related): add support for the Exponential distribution.
CSV importation: improved handling of quote symbols.
Fix bug: incorrect standard errors when the --robust option is given for random-effects panel estimator in presence of missing values.
Fix bug: rename() function not working on Windows when the target filename already exists.
Fix Windows-specific bug: long strings could get "lost" in printf and sprintf.
Fix bug: on repeated estimation of a given arima model in a function one might get slightly different results.
Fix for Windows XP: produce a build of gnuplot that does not depend on API introduced in later Windows versions.
Fix for Mac: try to ensure the default font is not too big when printing a script.
Fix for text search on Windows: try to ensure the found text is actually scrolled into view.
GUI: enable a tabbed editor for foreign scripts.
GUI: fix breakage in function-package search facility.
GUI: fix crash on requesting "Numerical summary" for panel groups boxplot.
GUI: fix for GUI editing of MIDAS plots.
GUI: improve font-selection dialogs, including a "Reset to default" option.
GUI: don't show duplicates in the database browser window; in case of duplicates just show the newest version.
Build process: enable use of "configure ; make" when building gretl on Windows using MSYS2.