EViews.com

We've identified the issue. In the MacKinnon algorithm for the response surface, we have found that there are potential issues with the polynomial approximation to the response surface for extreme values. In this case, the statistic value is so out of bounds that the existing algorithm approximation...

I can't see an easy way to do that since we need linearity in the states.

You can just set the values of the C vector outside of the object.

Sorry, the option was not documented. If you go into the dialog and run it with the additive outlier and look at the command capture, you'll see that the option is "type=ao", as in


Sorry for the inconvenience.

The 99% was a generic comment, not specific to your data (which I haven't seen). What I'd recommend is that you make a model from your equation and solve for the fitted probabilities and examine them. I do note that your 0 category is the least frequent, and though you most certainly have fitted pro...

Thanks. We will take a look.

The model is what it is. If you have a better model, then you can use it instead. Adding explanatory variables may help, but that may lead to overfitting. And in a case where 99% of observations have the same response, there is not much that one case do. You may want to generate the individual respo...

In this type of model it is certainly possible that the highest predicted probability never matches the actual realization. This problem becomes more acute the greater number of possible outcomes in the speicfication. In this case, the highest fitted probability is always for the dep=1 category, und...

Bear in mind that the large sample properties of clustering-by-cross-section rely on lots of cross-sections, and the opposite for cluster-by-period. So your performance in allowing for between period correlation (clustering by cross-section) will be much better than in the other dimension. Intuitive...

I can't say without seeing the data. You are welcome to post the data here.

All that I know is that we check the effective number of observations in the fixed effects estimator. Again, my guess is that you have variables that are non-time varying.

In a general clustering framework it is possible for there to be multiple observations in the crossed dimension. In the panel framework we disallow this, but I wanted to make certain that I understood your data. On the one-way covariance estimators, the naming convention is the opposite of what you ...

It looks as though you have a single observation per period/cross-section. It also sounds as though you want to use the fixed effects tools. If you didn't need those, I would say that unstructuring the workfile and just estimating using the standard equation clustered standard errors would be the ea...

Sounds as though you don't have enough data. In particular, the error probably suggests that estimating the fixed effects model requires dropping regressors to account for singularities (i.e., there are non-time varying regressors).

So there are multiple periods per firm-year? Have you already structured your workfile as a panel? Did EViews create a new sub-index?

For purposes of estimation, what panel features are you using? Lags, individual or period effects?

The panel estimators have built-in tools which allow for clustering by cross-section or by period, but not by both. So you can cluster by firm, and you can cluster by year, but not by firm-year. The non-panel estimators were recently updated to allow for arbitrary clustering. I am pretty sure that t...