Re: [Gretl-devel] The t variable

On Sun, 6 Nov 2011, Andreas Noack Jensen wrote: > I cannot get the internal t variable to work when data is yearly. > Here is an example > > # The t valible works when e.g. data is quarterly > nulldata 50 > setobs 4 1950:1 --time-series > series tmp = (t > "1960:1") > > # But not for yearly data > setobs 1 1950 --time-series > series tmp = (t > "1960") > > Would it be easy to extend the behaviour to yearly data? I'm afraid that the answer to that is No. With any frequency other than annual data there is no ambiguity between a plain observation number such as "23" and a time-series index such as "1970:1" (first quarter of 1970) or "1970:01" (first month of 1970). But with annual data there is perfect ambiguity between "1664" meaning the 1664th observation and meaning the year 1664. Obviously, there are possible heuristics we could employ, but these would be subject to error.

I would suggest that a reasonable method is to define a variable named "year", as in genr year = time + 1950 and then use "year" rather than "t" in setting up series such as the "tmp" that you specify above.

Allin Cottrell _______________________________________________ Gretl-devel mailing list Gretl-devel(a)lists.wfu.edu http://lists.wfu.edu/mailman/listinfo/gretl-devel