AudioBanter.com - View Single Post

Bill Riel wrote:
In article , says...

I have a quick question...

I need a numerical approximation to Student's t distribution that is
accurate to double precision.

Unless I am completely mistaken, the TDIST() function in Excel behaves
very strangely.

It cannot calculate a value for x 0 and the result of this function is
the pdf of the distribution, not the cdf unlike NORMSDIST().

I realize that I could construct a better function using the BETADIST()
function, but I'm not sure whether this will alleviate or compound
whatever inaccuracies are already present in Excel.

Currently, the one I am leaning towards is Numerical Recipes.

I don't know about Excel - I do recall that in the past that some of the
statistical functions were questionable.

However, I've used the code from Numerical Recipes and it's worked well
- it should meet your precision requirements.

Thanks, Bill. That's exactly what I did - a bit convoluted given that
you needed a Log Gamma and Incomplete Beta integral in order to get the
t, but probably worth it given the precision I am striving for.