R help - Jun 2006 - lme - Random Effects Struture

Thanks for the help Dimitris,

However I still have a question, this time I'll be more specific,

the following is my SAS code



proc mixed data=Reg;
        class ID;
        model y=Time Time*x1 Time*x2 Time*x3 /S;
        random intercept Time /S type=UN subject=ID G GCORR V;
        repeated /subject = ID R RCORR;
run;    **

(Type =UN for random effects)



The eqivalent lme statement I am using is :

reglme  <- lme(y ~ Time+Time*x1+Time*x2+Time*x3, data=Reg, random = ~ Time |
ID)



When I compare the results, the values differ by considerable margin; I
suppose this is due to the Random effects covariance structure. R output
tells me that the structure is



"Structure: General positive-definite, Log-Cholesky parametrization"



Hence the problem for me is how to control this structure in R. Any help
would appreciated

Thanks

Harry


On 6/28/06, Dimitris Rizopoulos <dimitris.rizopoulos@med.kuleuven.be>
wrote:
>
> AFAIK lme(), by default, estimates the covariance between two or more
> random-effects, thus you do not have to specify anything. You're
> probably looking for something like:
>
> fit <- lme(y ~ time * treat, random = ~ time | subject)
> fit
> summary(fit)
>
>
> I hope it helps.
>
> Best,
> Dimitris
>
> ----
> Dimitris Rizopoulos
> Ph.D. Student
> Biostatistical Centre
> School of Public Health
> Catholic University of Leuven
>
> Address: Kapucijnenvoer 35, Leuven, Belgium
> Tel: +32/(0)16/336899
> Fax: +32/(0)16/337015
> Web: http://med.kuleuven.be/biostat/
>     http://www.student.kuleuven.be/~m0390867/dimitris.htm
>
>
> ----- Original Message -----
> From: "harry wills" <harry.wills@gmail.com>
> To: <r-help@stat.math.ethz.ch>
> Sent: Wednesday, June 28, 2006 9:02 AM
> Subject: [R] Linear Mixed Effects
>
>
> > Hi,
> > I have implemented the lme in SAS and specified the random effects
> > covariance structure as unstructured using the statement "random
/
> > type=UN"
> > I want to specify same in R but not able to know how to do it.
> > Can anyone please advise me on how to proceed.
> > Thanks
> >
> >
> > --
> > Wills, Harry
> >
> > [[alternative HTML version deleted]]
> >
> > ______________________________________________
> > R-help@stat.math.ethz.ch mailing list
> > https://stat.ethz.ch/mailman/listinfo/r-help
> > PLEASE do read the posting guide!
> > http://www.R-project.org/posting-guide.html
> >
>
>
> Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm
>
>

-- 
Wills, Harry

	[[alternative HTML version deleted]]

On Wed, 2006-06-28 at 11:04 -0400, harry wills wrote:
> Thanks for the help Dimitris,
> 
> However I still have a question, this time I'll be more specific,
> 
> the following is my SAS code
> 
> 
> 
> proc mixed data=Reg;
>         class ID;
>         model y=Time Time*x1 Time*x2 Time*x3 /S;
>         random intercept Time /S type=UN subject=ID G GCORR V;
>         repeated /subject = ID R RCORR;
> run;    **
> 
> (Type =UN for random effects)
> 
> 
> 
> The eqivalent lme statement I am using is :
> 
> reglme  <- lme(y ~ Time+Time*x1+Time*x2+Time*x3, data=Reg, random = ~
Time |
> ID)
> 
> 
> 
> When I compare the results, the values differ by considerable margin; I
> suppose this is due to the Random effects covariance structure. R output
> tells me that the structure is
> 
> 
> 
> "Structure: General positive-definite, Log-Cholesky
parametrization"
> 
> 
> 
> Hence the problem for me is how to control this structure in R. Any help
> would appreciated
> 
> Thanks
> 
> Harry
>From my understanding of SAS, a*b means the interaction of a and b. But
in R, a*b is shorthand for a + b + a:b where a:b is the interaction
term. The way you've written the lme formula, you have time showing up 4
times plus you have additional main effects x1, x2, and x3. Is this what
you want? Maybe I'm wrong but I don't think the SAS code and the R code
represent the same model.

Rick B.