How to obtain p values when estimation with t distribution

[strypste]

Hi All

I wondered how Eviews calculates the p values if it estimates a model with maximum likelihood (the logl object) and the error distribution is assumed to be a t distribution (the degrees of freedom are also estimated by ml).

Thanks for your help!

Best
s

[EViews Glenn]

The error distribution is generally divorced from the distribution of the coefficients as we usually assume asymptotic normality of an appropriately centered version of the coefficients.

[strypste]

Hi EViews Glenn

Thanks for you reply. What do you mean with "appropriately centered version of the coefficients"? Or if this is too complicated to explain here, do you have any references on this issue?

Cheers
Steven

[EViews Glenn]

Centered as in the difference between the coefficient and the true parameters. Asymptotic normality will be discussed in almost any textbook that discusses maximum likelihood. Davidson and MacKinnon, Wooldridge, etc.