Page 1 of 1
Likelihood ratio
Posted: Tue Mar 11, 2014 1:34 pm
by startz
The following looks like a bug to me. I've done a redundant variable test following a probit (first attachment). The LogL values do not match the LR in the restricted model, nor in the original model (second attachment). In addition, I think the LR stat is off by a factor of two. Or maybe I'm misreading something???
- Redundant variables test
- Clipboard01.png (34.81 KiB) Viewed 6044 times
- Original estimate
- Clipboard02.png (27.66 KiB) Viewed 6044 times
Re: Likelihood ratio
Posted: Wed Mar 12, 2014 12:59 pm
by EViews Glenn
Workfile, please?
Re: Likelihood ratio
Posted: Wed Mar 12, 2014 1:24 pm
by EViews Glenn
You're missing something :)
The LR statistics are the ML equivalent of the F-statistics for the equation (against the base - constant only model). You want to be looking at the "Log likelihoods" which are used to populate the Restricted and unrestricted LogL rows of the Redundant Variables Test.
Re: Likelihood ratio
Posted: Wed Mar 12, 2014 1:28 pm
by startz
You're missing something :)
The LR statistics are the ML equivalent of the F-statistics for the equation (against the base - constant only model). You want to be looking at the "Log likelihoods" which are used to populate the Restricted and unrestricted LogL rows of the Redundant Variables Test.
Maybe could you add a head smacking emoticon to the forum selection??? I think I need one.
Re: Likelihood ratio
Posted: Wed Mar 12, 2014 2:08 pm
by EViews Gareth
I'll be happy to drive up the 5 and smack you in person.
Re: Likelihood ratio
Posted: Wed Mar 12, 2014 2:11 pm
by startz
I think the emoticon is supposed to show one smacking one's own head. But if you would like to drive up I will gladly supply you one or more beers as a peace offering.