Problem with copula regression

[Fabian]

Hi,

I have a dataset with a number of variables, which I'll call y1, y2, x1, x2, x3 and x4. y1 and y2 are both binary. I want to first run two regressions like:

y1 = x1 + x2 + x3 + x4
y2 = x1 + x2 + x3 + x4

Then, I want to run a number of coupla regression models (mainly FGM, Frank and Gaussian) to test the dependence between the y variables.

My understanding is I need to program these copula functions manually with ado files, using something like this:

http://www.stata.com/features/overvi...od-estimation/
http://www.stata.com/stata13/maximum...t-programming/

but after a lot of googling and searching, I've been unable to figure out exactly how to do it. I would be very appreciative if I could get some help or pointers on how to accomplish this.
Thank you very much,
Fabian

[P Van Kerm]

You need to write your likelihood evaluator indeed---there are many resources you can learn from. On the copula theme, I'd suggest you look at this recent Stata Journal article (and accompanying ado files): http://www.stata-journal.com/article...article=st0308 . It may get you on track.

Philippe

[Fabian]

Thanks for the quick reply Philippe. That link was one of the many I looked at, but I couldn't make much sense of how it worked exactly, and how I needed to modify it to do what I want to do. I'll keep trying, but I'm afraid I might be in over my head slightly

Fabian