Hi,I'm curious to set up the recursion-based model for a STCDO descibed by Andersen, et al (heterogeneous portfolio) and Gibson (homogeneous portfolio) using a Student t copula instead of the Gaussian copula and I had the following questions:1. Can anyone suggest a place where I can find a good algorithm (or code) for the cumulative t distribution? I believe TDIST() is meant to be pretty horrible and the one described in Numerical Recipes is not much better.2. I also need a suggestion for the inverse cumulative distribution, as I believe TINV() is not much better.3. Lastly, for the cumulative distribution what assumption of degrees of freedom is recommended? I recall reading in either the Andersen or Hull-White paper that they got good results with 4 or 5. I was wondering whether this is based more on experimentation, or whether there is a science to it.Thanks in advance.

Andersen's paper has a method for generating the required distribution and rarely needs more than 2 degrees of freedom, start out with one degree of freedom and increase until you match the constraints on the matrices given by andersen. Also check out shaw's paper on student-t, he's a prof in the maths dept at king's college, london

Hi Wibble,I actually have Gibson's LPM (Large Homogenous Pool) set up in a spreadsheet somwhere using the Gaussian copula. I'd like to transform this to a Student t copula.However, I don't think it's as straightforward as substituting NORMSINV() and NORMSDIST() with their corresponding T functions. I think the factors themselves have to be scaled too, right?I recall a similar thread on Wilmott from a while back. Lots of ideas were tossed around, but no conclusions reached.

Actually you just need the inverse of a student. I would say that a simple way to find it by approximation is by doing a simulation of student rv and numerically inverting it.You will need also the density of a gaussian and the density of a chi square (which has the same number of degree of freedom as your student).

there is a nice piece on the tdist in excel here As for pdf etc there is stuff on the web.