R help - Nov 2011 - SAS to R: I would like to replicate a statistical analysis performed in SAS in R.

Hello everybody, 


A statistician performed an analysis in SAS for me which I would like to
replicate in R.
I have however problems in figuring out the R code to do that.


As I understood it was a "covariance regression model". In the
analysis, baseline was used as covariate and autoregressive (1) as covariance
structure. The model included baseline, session, group and interaction between
baseline and group. The degree of freedom  was corrected by Kenward-Roger
method.


Study design:
- 2 Groups: group A: n=10 subjects, group B: n=10
- 5 motor test: baseline; learning sessions immediately, 6h, 24h, and 30d after
memory task
- dependent variable: response time (RT)
Research question: Do the two groups learn differently across time?


I would appreciate a lot any help on the R commands or the choice of statistical
approach.


Thanks!
Marianne, Columbia, SC

 		 	   		  
	[[alternative HTML version deleted]]

Marianne Stephan <mariannestephan <at> hotmail.com> writes:
> A statistician performed an analysis in SAS for me which I would
> like to replicate in R.  I have however problems in figuring out the
> R code to do that.  As I understood it was a "covariance regression
> model". In the analysis, baseline was used as covariate and
> autoregressive (1) as covariance structure. The model included
> baseline, session, group and interaction between baseline and
> group. The degree of freedom was corrected by Kenward-Roger method.
> Study design: - 2 Groups: group A: n=10 subjects, group B: n=10 - 5
> motor test: baseline; learning sessions immediately, 6h, 24h, and
> 30d after memory task - dependent variable: response time (RT)
> Research question: Do the two groups learn differently across time?
> I would appreciate a lot any help on the R commands or the choice of
> statistical approach.  Thanks!  Marianne, Columbia, SC [[alternative
> HTML version deleted]]
  You would be better off submitting this to the r-sig-mixed-models
mailing list.  If possible, posting your data somewhere, or a reasonable
facsimile of it, would improve your chances of getting an answer.
Kenward-Roger has not been available in R until recently
(see a long thread referenced at http://glmm.wikidot.com/faq on
"degrees of freedom", but has recently been implemented in
(an experimental version of?) the 'doBy' package.

   You might also get more answers if you show what you have
tried so far.

  Ben Bolker