From nwang@genie.tamu.edu  Fri Aug 16 00:38:09 1996
Received: from genie.tamu.edu (GENIE.TAMU.EDU [165.91.112.236]) by helix.nih.gov (8.6.13/8.6.12)
	 with ESMTP id AAA27568 for <rayc@helix.nih.gov>; Fri, 16 Aug 1996 00:38:08 -0400
Received: (from nwang@localhost) by genie.tamu.edu (8.7.5/8.7.3) id XAA04791; Thu, 15 Aug 1996 23:37:58 -0500 (CDT)
Date: Thu, 15 Aug 1996 23:37:58 -0500 (CDT)
From: Naisyin Wang <nwang@genie.tamu.edu>
Message-Id: <199608160437.XAA04791@genie.tamu.edu>
To: bobby@genie.tamu.edu, rayc@helix.nih.gov, xlin@sph.umich.edu
Cc: nwang@genie.tamu.edu
Subject: data
Status: R


Hello, everybody,

I tried  several different subsamples,  except the one (data set A)
of subjects with at least one abnormal ECG result (which gives \theta around 
0.4), the rest all give  \theta between 2 and 3.  These data sets include 
(i) people with the 1st ECG test abnormal (n=101) (ii) people with 
CHD = 1 or 2 (n=101) and (iii) the union of the two sets (n=155).
The major difference between the original data and (i)-(iii) with 
(A) is that (A) rules out the people with completely zero (normal) 
ECG tests.  Apparently the large theta is due to the differences 
between people with completely normal ECG during the study period 
and the rest.

Xihong and I have discussed the problem with using data set A.
It doesn't seem to be an acceptable option. For the rest of
the sets, with such large \theta's, PQL and CPQL is not really 
reliable. Should we go for MLE?

Naisyin