Jim Seward wrote:
----------------------------------------------------------------------------
I have some data for for subjects which includes repeated administrations
of the MMSE (a test of mental status that is usually considered a
continuous variable) and repeated lab values (see below);
ID Lab value a Lab value b MMSE
1 # # #
1 # # #
1
1 and so on...
2
2
2
3
3
3
3
4
4
I want to see if the MMSE scores vary in any systematic way with the lab
values. Note that there are some subjects with 4 observations, some with
3, and some with 2.
I essentially want to do a repeated measures multiple regression, with MMSE
as the outcome variable.
Is there such a thing?
----------------------------------------------------------------------------
Assuming that the two types of laboratory test results are not intolerably
correlated, how about -xtreg MMSE lab_value_a lab_value_b, i(ID) fe-? Based
on my understanding of the pitfalls of using random-effects (mixed-effects)
regression, I do not recommend -xt , re- for this, since I expect that the
laboratory test results will correlate with the random effect.
Graphical examination of the data is often helpful, and would seem
especially so here. Perhaps with a scatterplot matrix (-graph matrix MMSE
lab_value_a lab_value_b-), it will become evident if the MMSE scores
systematically vary with laboratory test results, and with time or sequence.
Joseph Coveney
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/