Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: uting for Non-independent observations using -xtgee- or -xtlogit-, pa???


From   "Wendell Joice" <wjoice@erols.com>
To   <statalist@hsphsun2.harvard.edu>
Subject   st: uting for Non-independent observations using -xtgee- or -xtlogit-, pa???
Date   Sun, 2 Mar 2003 20:12:04 -0500

Dear statalist,
Using stata 7, I am trying to perform a logistic regression analysis on a
dataset, covering several years, consisting of physician visit data and
various demographic or visit characteristic
variables. My basic unit (observation) is an individual patient visit to a
physician. Each visit consists of a different patient (no repeated visits
for any given patient) but each of the physicians has several patients.

(1) I am concerned about intra-physician correlation (non-independence of
observation violation) on the visits (possible correlation
between visits that occur to the same physician) and was
told that I could use either -xtgee- or -xtlogit, pa- to adjust for the
possibility of non-independence of observations  due to the fact that each
physician had several patients. Is this true?

(2) If true, I am testing whether or not certain variables
contribute to the patient receiving a psychotherapeutic treatment (yes, no).
I am confused as to whether the 'i()' option should refer to the individual
visit
variable or to the individual physician variable or what. So I tried many
possibilities such as the following and all resulted in the error message :
estimates diverging (absolute correlation > 1)
r(430);

psyment is a 1,0 dichotomous variable; visitid is individual visit; phycoden
is individual physician

For example, I tried the following using 'i()' as the individual visit
variable

xi: xtgee psyment i.agecatc i.sex i.met i.GEOGREGN i.hmo i.patseen
i.mntlrecx
      >  i.anyment psydinum i.cadg i.physpecy psydrnum  i.diagsrvcx  i.anxdr
i.psycotdr
      >  i.depresdr i.miscnsdr  i.ssricx i.phytimec i.insurtyp yearn if
csample==1,
      >  i(visitid) fam(bin) link(logit)

estimates diverging (absolute correlation > 1)
      r(430);

then I tried a simpler model,

 xtgee psyment psydinum psydrnum yearn if csample==1, i(visitid) fam(bin)
link(logit)

      estimates diverging (absolute correlation > 1)
      r(430);

then I switched to 'i()' referring to the individual physician variable

xtgee psyment psydin2 psydrnum yearn if csample==1, i(phycoden)
fam(binomial)
      >  link(logit) corr(exc) force

      Iteration 1: tolerance = .54993564
      Iteration 2: tolerance = .28833217
      estimates diverging (absolute correlation > 1)
      r(430);

then I tried xtlogit

. xtlogit psyment psydin2 psydrnum yearn if csample==1, i(phycoden) pa

Iteration 1: tolerance = .54993564
Iteration 2: tolerance = .28833217
estimates diverging (absolute correlation > 1)
r(430);

As you can see, none worked for me.
What am I doing wrong here?
Thanks
Wendell Joice




*
*   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/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index