[Gretl-users] Re: Create panel dataset

On Fri, 27 Sep 2019, Artur Tarassow wrote: "join" is not the only tool that could be used for this job but I think Artur is right: it should work well. Here's a little example, in the form of two scripts (which could be combined). The first reads some series from dbnomics -- annual population and employment for Germany, France and Italy -- and writes a time-series dataset; the second creates a panel dataset and fills it using "join", drawing from the time-series data. <hansl> # script 1: create time-series dataset set verbose off include dbnomics.gfn nulldata 38 setobs 1 1980 provider = "IMF" database = "WEO" bundle spec = null spec["weo-country"] = defarray("FRA", "DEU", "ITA") spec["weo-subject"] = defarray("LP", "LE") bs = dbnomics_get_multiple(provider, database, 20, 0, spec) list X = null loop i = 1..nelem(bs) --quiet sn = fixname(bs[i].series_code, 1) series tmp = NA dbnomics_bundle_get_data(bs[i], &tmp) # add series to list X += genseries(sn, tmp) endloop # save the per-country time-series data store tmp.gdt X </hansl> <hansl> # script 2: convert to panel # 38 years, 3 countries NT = 38 * 3 nulldata NT setobs 38 1:01 --stacked-time-series series country = $unit series Employment series Population strings cnames = defarray("DEU", "FRA", "ITA") # set the sample to each country in turn loop i = 1..3 --quiet smpl country == i --restrict --replace ename = sprintf("%s_LE", cnames[i]) pname = sprintf("%s_LP", cnames[i]) join tmp.gdt Employment --data="@ename" join tmp.gdt Population --data="@pname" endloop smpl full setinfo Employment --description="Employment - Persons" setinfo Population --description="Population - Persons" store dbn_panel.gdt </hansl> Allin Cottrell