Statalist


[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: RE: Triprobit & mvprobit errors (trivariate probit model)


From   "Tiffany Davenport" <tiffany.davenport@yale.edu>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: RE: Triprobit & mvprobit errors (trivariate probit model)
Date   Wed, 30 Jul 2008 10:36:47 -0400

Dear Stephen, 

Thank you very much for your comments. I appreciate your help. 

The sample size is 811. I specified the attempt at 50 because I received a
particular error at that point; I understood that the number of draws should
at least exceed the square root of the sample size (SJ 3(3)) but I have
tried larger values of R as well.  At 600 draws, I stopped the analysis
after hundreds of lines of "Warning: cannot do Cholesky factorization of rho
matrix." See syntax and output below.  I have also tried up to 3,400 draws
at which point nothing happened (no errors, no output) for 4 hours, and,
unfortunately, I could not leave the program running on this particular
occasion, so this may have been an issue of computing capacity. I am making
arrangements to run the larger number of draws on a different machine.  With
the -difficult- command, I receive the same ""Warning: cannot do Cholesky
factorization of rho matrix" and error 430 (syntax and output below).  I
will try some of the options for speeding up the computing time, reconsider
my specification, progressively increase the number of draws from a smaller
number upward, vary the technique, and look at your article in SJ 6(2) per
your suggestions.  Thanks again. 

Tiffany
(Output is below for your reference.)

50 draws:

. mvprobit (outcome= treatmentAcontact treatmentBcontact  g6 p6 g4 p4 g2 p2
stratum1) (treatmentAcontact= treatmentA g6 p6 g4 p4 g2 p2 stratum1)
(treatmentBcontact= treatmentB g6 p6 g4 p4 g2 p2 stratum1), cl(randunit)
dr(50)

Iteration 0:   log pseudolikelihood = -1061.8256  (not concave)
Iteration 1:   log pseudolikelihood = -754.05787  (not concave)
Iteration 2:   log pseudolikelihood = -672.60545  (not concave)
Iteration 3:   log pseudolikelihood = -574.53911  (not concave)
Warning: cannot do Cholesky factorization of rho matrix
Warning: cannot do Cholesky factorization of rho matrix

 [.~80 lines of this warning.]

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

600 draws:

. mvprobit (outcome= treatmentAcontact treatmentBcontact  g6 p6 g4 p4 g2 p2
stratum1) (treatmentAcontact= treatmentA g6 p6 g4 p4 g2 p2 stratum1)
(treatmentBcontact= treatmentB g6 p6 g4 p4 g2 p2 stratum1), cl (randunit)
dr(600)


Iteration 0:   log pseudolikelihood = -1061.8256  (not concave)
Iteration 1:   log pseudolikelihood =  -789.7645  (not concave)
Iteration 2:   log pseudolikelihood = -681.91457  (not concave)
Iteration 3:   log pseudolikelihood = -647.73038  (not concave)
Iteration 4:   log pseudolikelihood = -641.21325  (not concave)
Iteration 5:   log pseudolikelihood = -611.87758  (not concave)
Iteration 6:   log pseudolikelihood = -591.63921  (not concave)
Warning: cannot do Cholesky factorization of rho matrix
Warning: cannot do Cholesky factorization of rho matrix

[.~90 lines of this warning.]

Warning: cannot do Cholesky factorization of rho matrix
numerical derivatives are approximate
flat or discontinuous region encountered
Warning: cannot do Cholesky factorization of rho matrix

[.~80 lines of this warning.]

Warning: cannot do Cholesky factorization of rho matrix
--Break--
r(1);


200 draws -difficult-:

. mvprobit (outcome= treatmentAcontact treatmentBcontact  g6 p6 g4 p4 g2 p2
stratum1) (treatmentAcontact= treatmentA g6 p6 g4 p4 g2 p2 stratum1)
(treatmentBcontact= treatmentB g6 p6 g4 p4 g2 p2 stratum1), cl (randunit)
dr(200) difficult

Iteration 0:   log pseudolikelihood = -1061.8256  (not concave)
Warning: cannot do Cholesky factorization of rho matrix
Warning: cannot do Cholesky factorization of rho matrix

[.~70 lines of this warning.]

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


-----Original Message-----
From: owner-statalist@hsphsun2.harvard.edu
[mailto:owner-statalist@hsphsun2.harvard.edu] On Behalf Of Stephen P.
Jenkins
Sent: Wednesday, July 30, 2008 6:25 AM
To: statalist@hsphsun2.harvard.edu
Subject: st: Triprobit & mvprobit errors (trivariate probit model)
-----------------------------------------------
You haven't shown us EXACTLY what you typed or was in your do/log file
-- this is essential to ascertain EXACTLY what options you applied to these
commands. Nor do you say what the sample size is.

It's strange too that you report using only 50 draws for -mvprobit- but 5000
with -triprobit-. 

The problems appear to be with your data and model specification rather than
MSL.

The convergence problems you report suggest that. But there may be other
things that you have or have not tried.  E.g. with -mvprobit-, consider also
the difficult option, and also varying the technique option, e.g. something
like technique(dfp nr) can help.  

Also have a look at the Cappellari and Jenkins article in Stata Journal
6(2), with associated code, showing how you can speed up estimation of MV
Probit models using a plugin, and also exploiting Halton draws rather than
pseudo-random ones. (Preprint available as an
ISER working paper.)   On the choice of number of draws, see earlier
Statalist thread on this topic. Also see C + J article in SJ 3(3)
(downloadable from SJ website).    In short, once you get convergence,
you should explore what happens if you increase the number of draws ... keep
going until there is no change in the estimates to speak of.


Stephen
-------------------------------------------------------------
Professor Stephen P. Jenkins <stephenj@essex.ac.uk> Director, Institute for
Social and Economic Research University of Essex, Colchester CO4 3SQ, U.K.
Tel: +44 1206 873374.  Fax: +44 1206 873151.
http://www.iser.essex.ac.uk
Survival Analysis using Stata:
http://www.iser.essex.ac.uk/teaching/degree/stephenj/ec968/
Downloadable papers and software: http://ideas.repec.org/e/pje7.html

Learn about the UK's new household panel survey, the United Kingdom
Household Longitudinal Study: http://www.iser.essex.ac.uk/ukhls/ 


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

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



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index