a problem with LOGL object

[doaa]

Hi
I am using LOGL object to estimate my model but I got an error message. The model and the problem are written in details in the attached file. I apologize because you may find it a long one but I appreciate your help as I am in bad need to solve this problem as soon as possible

[trubador]

Could you please also upload your workfile that includes the variables? It may be helpful in tracking the estimation.

[doaa]

Hi
Enclosed herewith, please find the attached E-views file that includes data and model.

[trubador]

As you have suspected, you misspecified the likelihood. It seems you confused known E (errors from estimated GARCH model) with unknown R (residuals of the main equation inside the logl object). Although you used the former variable in generating other series, you should specify the latter in the loglikelihood equation. For instance, the conditional skewness should be defined as S = c(18)+c(19)*R(-1)^3+c(20)*S(-1). Residuals should be estimated jointly with these variables according to your parameter set. Other than that, your estimation difficulty seems typical of starting value problem. You can now try different initial values, once you make the necessary adjustments to your likelihood specification.

By the way, you can use @fact(x) function for factorial calculations...

[doaa]

Thank you very much for your effort and time. Actually, I found the peroblem is that I have not imposed the nonnegativity conditions on the parameters of variance, skewness, and kurtosis equations. Actually, I got the idea from your reply to one of the questions in the forum and I put the constraints and I got well-estimated model

[trubador]

Glad to hear that. However, I still believe you should revise your likelihood specification, since it does not match to that of original model in the documentation you sent.

[doaa]

I am really grateful for advise and I'll take it in my consideration for sure.