Christopher W. Ryan:
You would want to cluster on the medical record number, assuming that
is a unique patient ID, since you've probably got several varieties of
serial correlation, but that does not help with the variable to put on
the right hand side in your regression--I don't think you want visit
number, but maybe date of visit measured in days since 1/1/1960 or
somesuch (see -help dates- for more info) would be appropriate. Try:
. regress bmi visitdate, cluster(mrn)
. xtreg bmi visitdate, cluster(mrn) fe
. xtreg bmi visitdate, cluster(mrn)
Then you might want to see if the growth rate varies by initial BMI,
by interacting initial BMI with date, or cutting the sample. But
ultimately you might want to use -xtmixed- to model the growth rates,
which models growth rates as individual-specific by making some
assumptions about the distribution.
--Austin
On 5/11/06, Christopher W. Ryan <[email protected]> wrote:
I am trying to work with one of my residents who is studying obesity in
our practice. We have a dataset consisting of about 150 patients, who
collectively made about 700 visits over the course of two years. They
were selected for study because their BMI > 30. Number of visits per
person ranges from 1 to over 15.
-regress bmi visitnumber, cluster(mrn)-
Am I on the right track with this? I have never used -cluster()-
before. I won't be learning anything about time series analysis until
the fall, and I wonder whether it might be appropriate in this situation.
*
* 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/
