Multivariate probit

[Ismail Adam]

Dear All,

I got this error messeng when I run :

mvprobit (select1 = Age EDUC Market1a BGRADE EXPREJ Formality duration_ BPOWER FCONT) ///
(select2= Age EDUC Market1a BGRADE EXPREJ Formality duration_ BPOWER FCONT) in 176/377



Iteration 0: log likelihood = -417.41925 (not concave)
Warning: cannot do Cholesky factorization of rho matrix
Iteration 1: log likelihood = -351.89714
Warning: cannot do Cholesky factorization of rho matrix
Warning: cannot do Cholesky factorization of rho matrix

[Joro Kolev]

Sounds to me like an ill conditioned data. Why are you using the restriction - in 176/377-? Does the model go through if you do not restrict the sample?

[Hong Il Yoo]

It would also help us narrow down the list of suspects if you try estimating the same model using -biprobit- & share with us what you find.

[Ismail Adam]

Dear Professor. @Joro Kolev

Many thanks for your reply.

the total sample size is 430 actors (collector, traders, and processors), and I would like to do a multivariate probit for traders only (from 176 to 377)

yours sincerely
Ismail

[Ismail Adam]

Dear Professor. @Hong Il Yoo,

Many thanks for your reply.

actually, I have four dependent variables (select1, select2, select3, and select4)

could you please help me with this problem?

Best Regards,
Ismail

[Ismail Adam]

mvprobit (select1=Age EDUC Market1a BGRADE EXPREJ Formality duration_ BPOWER FCONT) (select2=Age EDUC Market1a BGRADE EXPREJ Formality duration_ BPOWER FCONT) (select3=Age EDUC Market1a BGRADE EXPREJ Formality duration_ BPOWER FCONT) (select4= Age EDUC Market1a BGRADE EXPREJ Formality duration_ BPOWER FCONT)


Note: 7 failures and 0 successes are completely determined.

Iteration 0: log likelihood = -601.33513



Warning: cannot do Cholesky factorization of rho matrix
Warning: cannot do Cholesky factorization of rho matrix
Warning: cannot do Cholesky factorization of rho matrix
Warning: cannot do Cholesky factorization of rho matrix
Warning: cannot do Cholesky factorization of rho matrix
Warning: cannot do Cholesky factorization of rho matrix
Warning: cannot do Cholesky factorization of rho matrix
could not calculate numerical derivatives
flat or discontinuous region encountered
r(430);

[Hong Il Yoo]

Ismail Adam I haven't used the command before, but from -help mvprobit-, we can learn that:

"In difficult maximization problems, the message "Warning: cannot do Cholesky factorization of rho matrix" may appear between iterations. It may be safely ignored if the maximization proceeds to a satisfactory conclusion. Results may differ depending on the sort order of the data, because the sort order affects which values of the random variable(s) get allocated to which observation. (Be assured, however, that mvprobit does not change the sort order of the data.) This potential problem is reduced, the larger the number of random draws that is used."

Based on this, one thing you can check is whether the use of more draws helps. For instance, you can try -mvprobit (...), draws(100)-. You can also experiment with whether using different sets of draws helps by changing the random number seed, for example -mvprobit (...), draws(100) seed(890801)-. Note that your current estimation run implicitly applies -draws(5)- and -seed(123456789)-.

Having said all that, if I were you, I would double-check whether my data and component model specifications are in order, before estimating the full four-equation model. For instance, you may want to check whether (1) you can estimate separate -probit- models for each equation and (2) you can estimate -biprobit- models for at least some of possible pairs of equations.

[Ismail Adam]

@Hong Il Yoo many thanks for helping me I tried biproit it works well

[abdi hussen]

dear ismail, i also encountered similar problem. did you get solution to "
Warning: cannot do Cholesky factorization of rho matrix"? if yes please share with me thanks.

[abdi hussen]

dear all please those who know Stata command for "iterative maximization" help me thanks