Re: NAN handling in statistics functions

I was thinking of the Cauchy distribution, with undefined variance. But Augustin-Louis Cauchy had quite a few things named after him. I know best Cauchy sequences as a construction of Real numbers. On Sun, Aug 29, 2021, 2:36 AM Stephen J. Turnbull < stephenjturnbull@gmail.com> wrote:
David Mertz, Ph.D. writes:
I have a distribution for you: Cauchy. :-)
Oh? Because NaN is the *result* of `stats.variance(cauchy)`?
It still seems like as INPUTS to a stats function NaN ~= missing.
Well, that was a little opaque. I wanted to fit it into the "I have a ... for you" pattern.
What I meant by "Cauchy" was just a ratio whose denominator's support includes zero. It's occasionally seen in (not so hot) dynamic economic trade models (designed on the assumption of positive growth rates oops :-�)P). I don't know about other contexts.
Steve
participants (1)
-
David Mertz, Ph.D.