On Thu, 26 Oct 2006, Ignacio Díaz-Emparanza wrote: Thanks for the report. I've fixed the bug in CVS and will make a new Windows snapshot available shortly. Two other recent fixes: * When restricting the sample using a boolean criterion in the GUI program, in the case where the sample is already restricted, the option to replace the current restriction got broken at some point. That's now working again. * I noticed that the Ljung-Box statistic you get as part of the autocorrelation test output for a model was not the same as the statistic you get if you save the residuals and do the correlogram "manually". (This is after the recent fix for the latter.) This was due to a difference in the methodology for calculating the ACF in the two cases. I have now made this consistent. -- Allin Cottrell Department of Economics Wake Forest University, NC