If I understand you correctly:
You should regress Y on X1 and X2, and cluster by ED to allow for
arbitrary serial correlation. Include a time trend to allow the usage
of Y to change over time, and include a dummy for the quality
initiative started at month 7. I hope it didn't start in every Dept
in month 7 (i.e. I hope you have some kind of control group).
#delimit;
g byte y=(MedicationY=="Yes");
g byte smoker=(Smoker=="Y");
g tm=ym(year(date(substr(var1,1,3)+" 15
"+substr(var1,4,2),"mdy",2060)),month(date(substr(var1,1,3)+" 15
"+substr(var1,4,2),"mdy",2060)));
format tm %tmYM;
qui su tm;
gen month=tm-r(min)+1;
probit y smoker age month qualinit, cluster(EmerDept);
If you have sampling weights, the problem is slightly different.
-----Original Message-----
From: [email protected] [mailto:[email protected]]
Sent: Friday, July 08, 2005 3:36 PM
To: [email protected]
Subject: st: repeated sampling of emergency department patients
We have the following dataset:
Monthly survey of patients who report to an emergency department (across
15 months) and if they use medication Y (y/n). We also have other data
on the patients' characteristics (X1, X2, etc).
6 different hospital emergency departments were surveyed. Each month we
targeted 15 patients, but sometimes less (or more) were enrolled.
ID Emer Dept Date Smoker Age Medication Y
1 1 jan03 Y 10 Yes
2 1 jan03 N 45 No
3 2 feb03 Y 23 No
We are trying to assess whether:
1) Use of Y changes across time
2) If X1, X2 are predictors of use of Y
3) Use of Y is associated with a particular emergency department
4) A quality initiative started at month 7 makes a difference in the use
of Y
*
* 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/
