SR Millis wrote:
Thanks, Joseph. I've resolved the discrepancies. The
most obvious problem was a data entry error on my
part. When I corrected my error, I obtained the same
results.
However, the pkcross output for "treatment effect" and
"period effect" seems to be the exactly converse of
the SAS output when using code by Senn ("cross-over
trials in clinical research") and by Brown & Prescott
(applied mixed models in medicine). That is, what is
"treatment" for Stata is "period" for S and B&P--and
what is "period" for Stata is "treatment" for S and
B&P.
----------------------------------------------------------------------------
It shouldn't be necessary to do data entry: just -use- the Chow and Liu
dataset from StataCorp's Web site and then -outsheet- to where SAS can find
it. You should be able to pull the dataset into SAS as the delimited ASCII
file.
In any event, I recommend that you check your coding. I have the first
edition of Chow and Liu,* and Stata's results agree with the authors' (see
Tables 3.6.2 and 3.6.3 on Pages 68 and 69). Note that they refer to the
sequence effect as the carryover effect. In this case, the terms are
interchangeable.
If you wish to reproduce the results shown in the first edition of Chow and
Liu, be sure to include the sequence (carryover) predictor in the model.
Brown and Prescott** do not in their examples of the AB / BA design (Tables
7.2 and 7.3 on Pages 266 and 267, see also discussion on Page 294), and,
although I don't have his book, Stephen Senn undoubtedly doesn't as well.
Joseph Coveney
* S-C. Chow and J-P. Liu, _Design and Analysis of Bioavailability and
Bioequivalence Studies_ (New York: Marcel Dekker, 1992)
** H. Brown and R. Prescott, _Applied Mixed Models in Medicine_ (Chichester:
John Wiley & Sons, 1999)
*
* 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/
