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

Re: Re: st: RE: ancova?


From   Joseph Coveney <jcoveney@bigplanet.com>
To   Statalist <statalist@hsphsun2.harvard.edu>
Subject   Re: Re: st: RE: ancova?
Date   Fri, 21 Feb 2003 00:45:26 +0900

Diachi Nozaki asked about a factorial randomized blocks ANCOVA as a possible 
approach to a set of experimental data, and Al Feiveson and I responded.

It dawned on me that there is an error in my posting.  I wrote that one degree of 
freedom must be deducted from the ANOVA table in order to account for the use of the 
covariate.  Instead, I believe that one degree of freedom must be deducted from the 
model in Step 2 for each degree of freedom consumed in the regression in Step 1.  If 
Daichi opts for a fixed-effects regression (-xi: regress response i.subject*assocvar-), 
then there are 19 degrees of freedom for 20 parameters estimated:  one intercept for 
Subject 1 and one difference from it for each of the other nine subjects and one slope 
for Subject 1 and nine differences from it.  If Diachi goes with random-coefficients 
regression (-generate byte k=1-, -eq intercept: k-, eq slope: assocvar-, -gllamm 
response assocvar, i(subject) nrf(2) eqs(intercept slope) adapt trace allc-) then there will 
be four or five parameters estimated:  one population mean intercept and one variance 
about that population mean, one population mean slope and one variance about that, 
and optionally one covariance parameter (-gllamm- estimates the covariance by 
default, but it can be toggled off).

Joseph Coveney
*
*   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