Thanks, Martin. Your comments are helpful. But I want to know about baseline adjustment in the specific case of random intercept/coefficient models -- say, once the decision has been made that a random coefficient model is the appropriate choice (e.g. a growth-type model or a change model). The reason I ask is that at a presentation of such a model for a small clinical trial, I questioned why a baseline factor was not in the model for change over time. The response was that baseline adjustment is implicit in the random intercept/coefficient model, so it is not necessary. I can see that variance components show the individual variability around the various population estimates, but I'm not sure that "adjustment" is accomplished. Maybe I'm trying to understand it from an ANOVA approach and this isn't the right way to look at it.
________________________________________
From: [email protected] [[email protected]] On Behalf Of Maarten buis [[email protected]]
Sent: Friday, November 13, 2009 3:59 PM
To: [email protected]
Subject: Re: st: baseline adjustment in mixed models
--- On Fri, 13/11/09, Visintainer PhD, Paul wrote:
> I have a question about baseline adjustment . . .
>
> Do random intercept models (RI) or random coefficient
> models (RC) account for group differences in the baseline
> value of the outcome? I'm not asking the general
> question of whether or not we should control for baseline
> values (there is a lot of literature on this), but rather,
> with RI or RC models is it even necessary?
>
> Suppose in a clinical trial, where outcome is measured on
> multiple occasions over time, randomization did not achieve
> balance between treatment and control groups on the initial
> value (Y0), is it necessary to control for the baseline
> value of the outcome in a RI or RC model? Is it
> redundant? Does it improve precision or
> efficiency? (Let's assume the outcome is continuous,
> rather than categorical).
I have three thoughts on that:
1) A random intercept model can't account for confounding
variables, so whether or not your randomization was successful
won't make a difference in your choice whether or not to use
a random intercept model.
2) With repeated observations on the same subject you need
to take into account that these observations don't provide
as much information as the same number of observations on
all different subjects. You can ask me on 3 different days
whether or not I like chocolate, or you can ask 3 different
people whether they like chocolate. In the latter case you
have three pieces of information and in the former less than
3, and probably just 1. Random effects models are one way of
taking this into account.
3) If your dependent variable is not continuous then, even
if your experiment is perfect in every respect, your
estimate of the effect will not represent how an individual
can be expected to react to your treatment if you do not
control for the baseline. Rather you will than be comparing
the average outcome between groups defined by the explanatory
variables. (If I am allowed some shameless self-promotion,
there is an explanation, and further references, of this issue
here: http://www.maartenbuis.nl/dissertation/chap_7.pdf )
Hope this helps,
Maarten
--------------------------
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen
Germany
http://www.maartenbuis.nl
--------------------------
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
----------------------------------------------------------------------
CONFIDENTIALITY NOTICE: This email communication and any attachments may contain confidential and privileged information for the use of the designated recipients named above. If you are not the intended recipient, you are hereby notified that you have received this communication in error and that any review, disclosure, dissemination, distribution or copying of it or its contents is prohibited. If you have received this communication in error, please reply to the sender immediately or by telephone at (413) 794-0000 and destroy all copies of this communication and any attachments. For further information regarding Baystate Health's privacy policy, please visit our Internet web site at http://www.baystatehealth.com.
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
