Ordered probit estimation error message

[Joanne]

Hi all
I am estimating an ordered probit model, but it is giving me the following error message: non-positive liklihood value for observation 2006.
Considering that I am using a panel and there are 308 different 2006 observation per variable, I don't know where to begin looking for the problem. Any suggestions? thanks

[EViews Gareth]

Can you tell if this is at starting values, or during the iterative process?

[Joanne]

I'm not entirely sure. I'm used to working with time series data unfortunately and have never encountered such a problem. The message comes up after I click ok in the estimation window (with starting coefficent values set as eviews supplied). If I change the starting values to the zero option, then it allows me to estimate the regresstion, but if I set the starting coefficient values to any of the other options then it gives me that same error message.

[EViews Gareth]

You could stick with using zero, or, if you want to be more thorough, you could use user-supplied and try different combinations.