[R-sig-ME] Comparing random slopes?

Rafael Maia queirozrafaelmv at yahoo.com.br
Wed Jan 28 12:04:37 CET 2009

Dear list members,

I am analyzing a data set comprised of two variables measured over  
time (once a month, over a year, for a total of 12 observations per  
individual). I am interested in modeling the relationships between  
these two variables, but I'd like specifically to compare if  
individual variation over time is associated between them. There is  
great individual variation in slopes for both variables, and I'd like  
to measure the degree of intra-individual association between the  
slopes for the two.

My first intuition was to conduct random slopes models for the  
variables and compare random effects (conditional modes for the  
slope?). However, I recall reading in some slides from Douglas Bates  
that these are not actually "estimates", actually something more  
similar to predictors. I also think I remember someone mentioning on  
this list specifically that these values should not be extracted and  
used as parameters for other analyses.

If this was not a mixed-model "situation", I guess it would be  
reasonable to compare regression slopes through an analysis of  
covariance, for example. Anyone has any suggestion as to how to treat  
this case? Any help would be greatly appreciated.

Many thanks in advance!

Rafael Maia
"A little learning is a dangerous thing; drink deep, or taste not the  
Pierian spring." (A. Pope)
M.Sc., Animal Behavior Lab - Dept. of Zoology
Universidade de Brasilia, Brazil

More information about the R-sig-mixed-models mailing list