Hi Giovanni,
Yes, changing the nfirst value should change the results.
You are deciding on the first time series value to use in
the deconvolution procedure.
From a cursory glance at the code, it is not clear what
effect a smaller nfirst value would have. For stimulus
type with a maxlag value greater than the nfirst value,
one might worry that those first time points won't have
complete data for evaluation, e.g. with nfirst = 2, and
minlag/maxlag = 0/6, there are 3 time points missing in
order to fully evaluate time point 2.
It may be a good idea to use nfirst = 0, or it may not.
To me, either case may have a reasonable chance of being
correct. But I suspect only the author of the program,
Doug Ward, knows well enough to say.
- rick