[Gretl-users] system estimation on large cross-section problem under win XP (3)

Dear GRETL Folk, greetings from Paris ! I finally checked my problem and here are my conclusions: 1) in fact it is a data problem, because i have two series of prices that are essentially very colinear, thus the explanatory variables matrices are near singular but only so when i loop on the different years (for the whole cross-section the vif test show no colinearity vif <10.0 for all variables); 2) so, when i use ols commands into a loop of the kind : loop i=1..5 smpl SURVNUM=$i --restrict --replace ols WLOG ... ols WCAR_... end loop smpl full everything runs perfectly as expected, however, the ols results on my data clearly show that there is a colinearity problem: certain coefficients are too large in magnitude. More importantly, gretl tells me that some variables have been dropped because of colinearity and that the data matrices are still near-singular ! 3) A first line of conclusion is thus that : first, the loops are not under cause ; second, it seems that the system commands do not handle colinearity in the variables as efficiently as does ols. This seems rather serious because instead of getting an error message gretl crashes. I still work on the problem but i think it has been pinpointed. cheers & regards Franck -- Franck Nadaud CIRED UMR 8568 CNRS - EHESS, ENPC, ENGREF, CIRAD 45 bis avenue de la Belle Gabrielle 94736 Nogent-sur-Marne Cedex TEL: 33-1-43-94-73-94 FAX: 33-1-43-94-73-70 MOB: 06-07-39-92-75 France