Get Log Likelihood function equal to NaN when running VAR-estimation

Karl

2012 5 30

0 个回答

6 次查看（30 天）

显示更早的评论

0 个投票

Hi, I'm fairly new to Matlab, so this might be a stupid question. I'm currently trying to estimate a small, easy VAR-model for economic forecasting. At the moment I have six variables, and 73 (quarterly) observations in the 73x6 matrix Data.

I first define the model, and since I want four lags the code looks as follows:

Spec=vgxset('n', 6, 'nAR', 4, 'Constant', true);

I then estimate the model:

[EstSpec, EstStdErrors, LLF, W] = vgxvarx(Spec, Data(5:end,:),[],Data(1:4,:));

I then want to check if the number of lags is sufficient, but all such test need the loglikelihood function LLF. However, when doing my estimation the LLF is just a single NaN-entry. Any thoughts about why this is so, and what I can do about it?

Thanks in advance! Karl