Date: Wed, 2 Jul 2008 11:12:25 +0530
Reply-To: Madan Gopal Kundu <Madan.Kundu@RANBAXY.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Madan Gopal Kundu <Madan.Kundu@RANBAXY.COM>
Subject: Re: Model selection based on AIC in PROC MIXED
In-Reply-To: A<200807020529.m61LCKC9002951@malibu.cc.uga.edu>
Content-Type: text/plain; charset="us-ascii"
Hi Doug Robinson,
Does your model include any random effect? If not, then you can perform
'stepwise' or 'Forward' regression in PROC REG. In that case no need to
add variable one by one as you said in the trailing mail; SAS will do
that for you.
You said result of significant test (based on p-value) and
Log-likelihood test are not matching. Sometime it may happen if the
added variable contributes largely to the model, but may not appear as
significant due to the high variability.
In this case I would suggest you to calculate partial correlation of the
all the fixed variables with the dependent variable. Then keep adding
one by one variable to the model. It may help you.
Though, you find significant improvement using Log-likelihood test, I am
not in favor of keeping insignificant variables in the final model.
Regards
Madan Gopal Kundu
Biostatistician, CDM, MACR, Ranbaxy Labs. Ltd.
Tel(O): +91 (0) 1245194045 - Mobile: +91 (0) 9868788406
-----Original Message-----
From: SAS(r) Discussion [mailto:SAS-L@LISTSERV.UGA.EDU] On Behalf Of
Doug Robinson
Sent: Wednesday, July 02, 2008 10:59 AM
To: SAS-L@LISTSERV.UGA.EDU
Subject: Model selection based on AIC in PROC MIXED
Hi all, I'm using -2 Res Log Likelihood, AIC, and BIC values from PROC
MIXED
to help me chose a model that best fits my data on provisioning rates at
bird nests. I'm new to this technique and have a few questions that I
hope
you can help me with.
I've gone step-by-step and added terms to the model and noted the values
of
Fit Statistics, and their change with each addition to the model. I
know
how to test whether the addition of each term improves the model using a
Log
Likelihood Test, but what I'm confused about is the significance of the
model terms with respect to their 'p' values in the Type III Tests of
Fixed
Effects. If the model improves significantly (based on Log Likelihood
Test),
but the Type III p values indicate the variable does not explain a
significant amount of variance in the data (based on the p value), what
does
that mean? Do I stick with this model and leave the non-significant
variable in the model? Is there something else I should be examine
(residuals, etc.) to determine whether I'm missing an outlier or
something
similar?
I would appreciate any insights into my problem. Thank your for your
time
and attention.
(i) The information contained in this e-mail message is intended only for the confidential use of the recipient(s) named above. This message is privileged and confidential. If the reader of this message is not the intended recipient or an agent responsible for delivering it to the intended recipient, you are hereby notified that you have received this document in error and that any review, dissemination, distribution, or copying of this message is strictly prohibited. If you have received this communication in error, please notify us immediately by e-mail, and delete the original message.
(ii) The sender confirms that Ranbaxy shall not be responsible if this email message is used for any indecent, unsolicited or illegal purposes, which are in violation of any existing laws and the same shall solely be the responsibility of the sender and that Ranbaxy shall at all times be indemnified of any civil and/ or criminal liabilities or consequences there.
