Re: st: Fwd: -cmp- not concave

["Dimitriy V. Masterov" <dvmaster@gmail.com>]

Leo,

Your syntax seems nonstandard to me, particularly the plus sign,
though that doesn't seem to be the problem.

Does the user-written bitobit work by any chance?

DVM

On Thu, Mar 6, 2014 at 5:45 PM, Z. Liu <zl290@cam.ac.uk> wrote:
> Dear all,
>
> I've been trying to fit a bivariate tobit model using the user-written
> command -cmp-. The model looks like this:
>
> cmp (equation1: y1 = X1) (equation2: y2 = y1 + X2), indicators(2 2) robust
>
> This model doesn't work out. Stata keeps showing 'log pseudolikelihood = 0
> (not concave)'. I tried to start with minimum regressors (1-2 explanatory
> variables) in each equation, but the problem remains. I also tried
> alternative ml options such as - tech(dfp)- and -difficult- but still can't
> fix it. However, the two tobit equations can be estimated separately,
> although the Pseudo R2 is quite small (around 5%). As well, the system model
> can be estimated if I change the two tobit equations into linear equations.
> Some descriptive statistics of the two dependent variables are as follows:
>
> y1: Percentiles: 1% (0), 25% (0), 50% (23.57), 99% (6127.14); mean (365.00);
> SD (1186.90); Skewness (7.14); Kurtosis (71.23)
>
> y2: Percentiles: 1% (0), 50% (0), 75% (0), 90% (78.10), 99% (1128.53); mean
> (61.29); SD (401.82); Skewness (13.27); Kurtosis (217.49)
>
> Anybody out there knows what might be the problem? Any advice or suggestions
> would be appreciated. Thanks a lot in advance.
>
>
> Best Regards,
>
>
> Leo
>
>
> *
> *   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/
*
*   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/