Hope you are well and had a wonderful day.
I used MDCEV and MDCNEV models to model people's activities in one area. I use the code that Prof.Bhat puts on website in Gauss 10. The number of observations is about 15,000 people with 12 goods, with the effect of 15 independent variables being displayed. Each model's running time lasts about one day.
According to the code, I need to see the effect of each independent variable in each good in each run, which makes the whole term of the model's code implementation much.
The error I sometimes encounter when adding a variable to a good after the run code finished is "Hessian matrix was failed". Is there a way to avoid encountering this error before executing the code that will slow down this runtime?
thank you in advance for your time and consideration.
Take a look at the "Linear dependencies or nearly linear dependencies in the sampling distribution" section of this paper. This does not answer your question about how to know ahead of time, but it does help you find the cause of the problem.