[Gretl-users] Re: Split regression and Groupby Summary Statistics

Hi, regarding grouped summary statistics, see the built-in gretl function `aggregate()`: https://gretl.sourceforge.net/gretl-help/funcref.html#aggregate For computing multiple grouped statistics, see also the PandasPort package: https://gretl.sourceforge.net/current_fnfiles/PandasPort.gfn Example ``` set verbose off ### Use bult-in aggregate() function matrix mout_mean = aggregate(Targets, Groupby, "mean") print mout_mean matrix mout_median = aggregate(Targets, Groupby, "median") print mout_median ### USE PandasPort package # pkg install PandasPort # ONLY ONCE NEEDED FOR PACKAGE INSTALLTION include PandasPort.gfn open mrw.gdt --quiet # Aggregation for multiple methods strings methods = defarray("mean", "min", "max") # Groupy by single discrete series matrix result = agg(gdp85, OECD, methods) print result # Groupy by multiple discrete series list Targets = gdp60 gdp85 list Groupby = nonoil OECD matrix result = agg(Targets, Groupby, methods) print result ``` For regression this does not exist in gretl yet. Most simple solution I can imagine: Run a loop over the grouping variable and restrict the sample each iteration before running a regression. Artur