Hi Anderson,
I've noticed palm_gtoz function may produce Inf or -Inf values on Z and it causes NaN related errors afterwards. This happens when input of erfcinv is close to 0 or 2 within matlab's numerical precision. So, I would alternate Inf or -Inf values on Z as maximum or minimum values of Z to prevent this problem, by putting following codes in the end of palm_gtoz.m:
z_max = max(Z(Z ~= Inf));
z_min = min(Z(Z ~= -Inf));
Z(Z == Inf) = z_max;
Z(Z == -Inf) = z_min;
Does this make sense?
Thanks,
Dongjin
|