Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: log-likelihood comparison of logit, loglog and cloglog?
From
Alexander Kihm <[email protected]>
To
[email protected]
Subject
Re: st: log-likelihood comparison of logit, loglog and cloglog?
Date
Mon, 16 Sep 2013 14:46:55 +0200
Thanks a lot Nick.
I will indeed look at the other diagnostics as well, though it's a bit
more complicated as I am using "boxtid glm" to estimate the models
with predictor transformations.
The only thing I wanted to make sure is that one can actually compare
the log-likelihoods, as the models are somehow non-nested.
Alex
On 15 September 2013 08:12, <[email protected]> wrote:
> My short answer is Yes, it makes sense to look at log-likelihood, but also look at the usual output (z tests etc.), -glmcorr- (SSC). Also -findit modeldiag- for residual diagnostics (see the 2004 article there but download software from the update in 2010).
>
> Nick
> [email protected]
>
> On 14 Sep 2013, at 17:20, Alexander Kihm <[email protected]> wrote:
>>
>> my dependent variable is a proportion and I want to try three
>> different link functions when estimating my models with glm: logit,
>> loglog and cloglog.
>> Since I use the exact same predictors, I was thinking about comparing
>> the models' fit just by the log-likelihood.
>> Is that appropriate?
>>
>>
>
> *
> * For searches and help try:
> * http://www.stata.com/help.cgi?search
> * http://www.stata.com/support/faqs/resources/statalist-faq/
> * http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/faqs/resources/statalist-faq/
* http://www.ats.ucla.edu/stat/stata/