On Sun, 12 May 2013, artur tarassow wrote: > I've got an obvious issue with collinearity in my model. When I try to > compute IRFs with bootstrapped confidence intervals, gretl stops and gives > the following message: > "Excessive collinearity in resampled datasets". > > This is a good hint, no question. But what is the specific criteria for > gretl to stop here? Furthermore, despite of setting "set halt_on_error > off", gretl stops.
Here's a relevant comment in lib/src/irfboot.c: "In re-estimation of VAR or VECM we'll tolerate at most 9 cases of near-perfect collinearity (which can arise by chance): maybe this should be more flexible?" I'll see about respecting the "halt_on_error" setting. However, in the meantime "catch" can be used: catch matrix IR = irf(1, 2, 0.2) Allin
