On Sun, 16 Jan 2011, Allin Cottrell wrote:
> On Sun, 16 Jan 2011, Giuseppe Vittucci wrote:
>
> > I replicated the problem simply adding a (now useless) loop in
> > the new code:
> >
> > loop 5000 --progressive
> > dataset resample n
> >
> > #logit regression
> > logit A const bid X --quiet
> >
> > #mean and median Willingness-To-Pay
> > matrix x = {X}
> > scalar mean_wtp = - ($coeff(const)+(meanc(x)*$coeff[3:
> > $ncoeff]))/$coeff(bid)
> > scalar median_wtp = - ($coeff(const)+(quantile(x,0.5)*$coeff[3:
> > $ncoeff]))/$coeff(bid)
> >
> > #USELESS NESTED LOOP....
> > o = rows(x)
> > loop for i = 1...o --quiet
> > series s = 0
> > endloop
> >
> > #printing stats on these
> > print mean_wtp median_wtp
> > #save the coefficients to file
> > store wtp.gdt mean_wtp median_wtp
> > smpl full
> > endloop
>
> Thanks. I can confirm the segfault, and now that I have an example
> of the problem this should be fixed fairly quickly.
It turned out that the minimal script to provoke this bug looked
like this:
nulldata 5
loop 2
dataset resample 5
loop 1
series s = 0
endloop
endloop
The problem was specific to the case where an outer loop resamples
the dataset and an inner one defines a series. This should now be
OK in CVS and snapshots.
Allin Cottrell
