Re: [S] QUERY: why different $loglik[1] in survreg

Jens Oehlschlaegel (oehl@Psyres-Stuttgart.DE)
Tue, 24 Feb 1998 23:14:29 +0100 (MET)


On Tue, 24 Feb 1998, Barry Courtois wrote:

> Hi, I believe that if you are using na.action=na.omit and there are
> missing observations in the data (specifically the seven other predictors)
> you are fitting the value of the null-model likelihoods would be
> different. bnc

nice guess, but not in this case: no missings, same n. The models with
more predictors have lower $loglik[1] i.e. higher -2LL

the observed difference is

+3 for a log-normal
+4 for a identity-normal

on the -2LL scale

any other ideas?

> On Tue, 24 Feb 1998, Jens Oehlschlaegel wrote:
>
> > perhaps this is a dumb question: I fitted two nested models with
> > survreg(), one with 1 and one with the same and seven other predictors.
> > As expected I get different $loglik[2] at MLE, but also the $loglik[1]
> > differ. Shouldn't they represent a null-model and thus be equal? Is this a
> > related problem to the deviance problem recently discussed?

--
Jens Oehlschlaegel-Akiyoshi
Psychologist/Statistician
Project TR-EAT + COST Action B6
                                                 F.rankfurt
oehl@psyres-stuttgart.de                         A.ttention
+49 711 6781-408 (phone)                         I.nventory
+49 711 6876902  (fax)                           R .-----.
                                                  / ----- \
Center for Psychotherapy Research                | | 0 0 | |
Christian-Belser-Strasse 79a                     | |  ?  | |
D-70597 Stuttgart Germany                         \ ----- /
-------------------------------------------------- '-----' -
(general disclaimer)                             it's better

----------------------------------------------------------------------- This message was distributed by s-news@wubios.wustl.edu. To unsubscribe send e-mail to s-news-request@wubios.wustl.edu with the BODY of the message: unsubscribe s-news