PANEL ordered probit LR test missing

[Jayati Chatterjee]

Hi, I am trying to run a panel ordered probit model.
My iterations show as backed up, and all the variables have coefficients but the sigma2 u does not show the confidence interval.
the prob>chi2 is also missing
shows like this Prob > chi2 = .
Is this the reason I am not getting the marginal effects of this model?
My dependent variable has four categories. The first category has 52%, 2nd 5% , 3rd 13% and 4th 30 % of observations respectively.
How can I solve this?

The model runs with a different dependent variable but same set of independent variable or when I club the categories in two groups and run a probit model

[Jeff Wooldridge]

My opinion: Use oprobit and cluster your standard errors. This is more robust (in terms of consistency) and much easier computationally.

xtset id year
oprobit y x1 ... xK, vce(cluster id)

The margins command will work fine. My guess is there's a complicated serial correlation pattern that is leading to difficulties in estimating a single parameter. xtprobit is a joint MLE and is inconsistent in that case. oprobit allows any serial correlation pattern.

The only downside with oprobit is that it's not the most efficient estimator. But efficiency is not so desirable if the estimator is inconsistent.

[Jayati Chatterjee]

Thank you Sir. The oprobit is working fine, so the pooled model is better in this case I guess
I had tried earlier with xtmprobit that showed discontinuous region encountered.
I recognized my problem is with the industry specific control that I am using because MLE convergence occurs if I drop it (incase of mprobit y x1....xk, vce(cluster id)).