Re: comparing slopes for significant difference

dave_at_autobox.com
Date: 03/11/05 

Date: 11 Mar 2005 13:54:02 -0800

Ben,

Since you have correlated data ( observations over time on an
individual ) ..you can't simply use standard F tests as the effect of
correlated observations is to provide a FALSE estimate of the number of
observations.
Furthermore the underlying model for any one individual might well be

y(t)=y(t-1)+thetha0 for t=2,,3,4,...T

or some other autoprojective scheme

or

y(t)= constant for t=1,2,...t1

and

y(t)=some autoregressive function of the past of y + THETAO starting at
t2, and ending at T

Perhaps you might look around Oxford for someone who can help you with
time series data and time series forecasting.

Dave Reilly
AUTOMATIC FORECASTING SYSTEMS
http://www.autobox.com
215-675-0652 in the colonies