On Sat, 3 Oct 2009, Nikil van Wijk wrote: > Using a panel data set in Gretl, I'm trying to obtain a > F-statistic for country (fixed or random) effects and a > F-statistic for time (fixed or random) effects. Gretl offers the > possibility to generate both time and unit (country) dummies in > the 'add' menu, but in the 'model'-'panel'-'fixed or random > effects' menu, the only option is 'include time dummies'. (not > unit dummies). > > The statistic for the effect of country (dummies) cannot be obtained by > including all dummies in the dataset, because then they are 'Omitted > due to exact collinearity'. > > Any ideas on how to solve this?
Use gretl's fixed-effects estimator and the F-test for the country effects is printed automatically. Besides that, if you're puzzled by the message about exact collinearity, read up on dummy variables in any introductory econometrics textbook. Allin Cottrell
