Following up on yesterday's query, I decided to change the Output 5 assignment from previous years and NOT ask you to generate the person period data set for the modeling part of that homework.

I added the constant and the linear (centered) time function, but not the higher order time functions (you can do those).  A few additional notes:

a.        When you do the completely general time specification (with dummies), you'll get a wonky estimate at the final occasion. That's due to censorship.  Don't worry about it, since the point of the completely general model is less "interpretation" and more "estimate of best possible chisquare"

b.      Technically, the time specification here is wrong.  The real time spacing for the occasions in this data sat is 0, 0.08, 0.23, 1.23, 2.23, 3.23, 5.23.  Unfortunately, that time vector (even with centering) produces nutty estimates.  So I've entered the time function as 0,1,2,3,4,5,6 (or -3,-2,-1,0,1,2,3 after centering), because it produces better estimates.  You just have to be aware that the above vector appears to be imposing a linear function on equally spaced intervals when, in fact, that's not exactly what we have.


Michael Marsiske, Ph.D.
Associate Professor
University of Florida
Department of Clinical and Health Psychology
PO Box 100165, 1225 Center Dr., Rm. 3170
Gainesville, FL 32610-0165
[log in to unmask] (email)
(352) 273-5097 (phone)
(801) 720-5897 (fax) (website)