How to avoid inf

Dear all,

I take the log of difference between two standard normal CDFs,

that is

D=log(normcdf()-normcdf())

but in most of the cases D=inf. Is there a way to avoid having inf?

Thanks in advance,

Answer 1

[Udit Gupta]

Log of zero is -Inf. To avoid that add a small number before taking the log. For example -

    D = log(normcdf() - normcdf() + 0.00001)