Decomposition of Theil U?

DanielBachman · Wed Dec 12, 2007 11:17 am

I hope I am not missing something obvious, but I would like to know how to obtain the decomposition of Theil's U (into bias, regression, and residuals).

Any information anybody has about this would be very welcome.

TomDoan · Unread post by **TomDoan** » Wed Dec 12, 2007 4:53 pm

Theil's statistics for decomposition of forecast error aren't the same as the Theil's U statistics produced by the THEIL instruction. What I think you want are computed by the uforeerrors procedure with the theildecomp option. A couple of the Pindyck and Rubinfeld examples (pr08-2.prg and pr08-3.prg) use it.

DanielBachman · Thu Dec 13, 2007 9:06 am

Thank you for the prompt and clear reply. I knew you had to have the answer somewhere!