st: predicted probabilities with mnvp command

[Omar David Stabridis Arana <odstabridis@colmex.mx>]

My purpose is to estimate predicted probabilities after mvprobit. I have a trivariate probit and I revised Cappellari and Jenkins (2006) and one note of statalist wrote for Charlotte Berry (http://www.stata.com/statalist/archive/2009-07/msg00337.html) I followed and



mvprobit (y1= x1 x2) (y2=y1 x1 x3 ) (y3 = y2 x3 x4)

The one side, I used your instructions for estimate predicted probabilities similar predicted probabilities because mvppred only estimates when Y´s are one or are zero. In page 166 of Cappellari and Jenkins (2006) and Berry explain the steps for obtain other probabilities (001,010, 100, 110,101,011)  so I made this:

,mvppred stdp
correlate  stdp1 stdp2 stdp3
matrix V=r(C)
matrix c=cholesky(V)

gen k1=2*y1-1
gen k2=2*y2-1
gen k3=2*y3-1
mdraws, dr(500) neq(3) prefix(z) random seed(12345) antithetics replace
egen probab=mvnp(xb1 xb2 xb3) ,  chol(c) draws(500) prefix(z) signs(k1 k2 k3) adoonly

But, I have problems with this variable. It assumed that when I use mvnp  I get all 8 probabilities but I  doubt because when I use mvppred pall the estimations are all sample but my variable is only for many cases not all as mvppred all, my program is wrong?, Must I obtain 8 variables , one for each case? My variable probab is defined between zero and one, but if you write "bysort y1 y2 y3: sum probab" the sum of means for each case is upper to one.

The other side, I tried run the advice gave for Nick Cox to Charlotte "0. Before you call this code,  gen safecopy = .  1. Within your -myll- code add as last line  replace safecopy = `sp'  2. Access the variable -safecopy-afterwards". The problem is when I write sum safecopy the result is all missings, the observations in safecopy didn´t change. What is my mistake?

Thank you



*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/