s-news
[Top] [All Lists]

Re: Repeated-measures Designs

To: Jody.Biggs@csiro.au, s-news@lists.biostat.wustl.edu
Subject: Re: Repeated-measures Designs
From: "Richard M. Heiberger" <rmh@temple.edu>
Date: Tue, 31 Aug 2004 23:32:32 -0400
The breakout of the degrees of freedom for the ID stratum is an
artifact of the imbalance.  The ID stratum has 41 degrees of freedom,
as anticipated.  One df is for treatment.  With balance the remaining 
40 generated dummy variables of the ID effect would have been orthogonal
to the Hours and Treatment:Hours dummy variables.  Because there were dropouts,
the dummy variables for ID are not orthogonal to the dummy variables for
Hours and Treatment:Hours.  Since they are not orthogonal, they
project onto the linear space defined by the Hours and Treatment:Hours
dummy variables.

Rich

<Prev in Thread] Current Thread [Next in Thread>