R help - Apr 2013 - Mixed Modeling in lme4

Hi All,

I am trying to shift from running mixed models in SAS using PROC MIXED
to using lme4 package in R. In trying to match the coefficients of R
output to that of SAS output, I came across this problem.

The dataset I am using is this one:

http://support.sas.com/documentation/cdl/en/statug/63033/HTML/default/viewer.htm#statug_mixed_sect034.htm

If I run the following code:

proc mixed data=rc method=ML covtest;
class Batch;
model Y = Month / s;
random Int Month / type=cs sub=Batch s;
run;

The Fixed effect coefficients match with that of R. But the random
effect does not. Here is the R code:

rc <- read.table('rc.csv', sep = ',', header=T,
na.strings=".")

m1 <- lmer(formula = Y ~ Month + (Month|Batch), data = rc, REML = F)

summary(m1)

fixef(m1)

ranef(m1)

But if I change the SAS code as follows:

proc mixed data=rc method=ML covtest;
class Batch;
model Y = Month / s;
random Int / type=cs sub=Batch s;
run;

and the R code as follows:

m2 <- lmer(formula = Y ~ Month + (1|Batch), data = rc, REML = F)

summary(m2)

fixef(m2)

ranef(m2)

both fixed and random effect coefficients match. I am unable to
understand this discrepancy. Am I wrongly specifying the model in the
first case?

It would be helpful if someone can throw some light on this.

Regards,
Indrajit

Hi Indrajit,

In your first SAS code, change to type=un.  cs imposes the (somewhat
dubious) assumption that the variance of both the intercept and slope are
equal.  If you are using lme4, all random effects in a single block (e.g.,
(1 + month | batch) the 1 = intercept and month = random slope) will have
an unstructured (or freely estimated) variance covariance matrix.

Cheers,

Josh



On Mon, Apr 29, 2013 at 11:26 PM, Indrajit Sengupta <
indrajitsg2013@gmail.com> wrote:
> Hi All,
>
> I am trying to shift from running mixed models in SAS using PROC MIXED
> to using lme4 package in R. In trying to match the coefficients of R
> output to that of SAS output, I came across this problem.
>
> The dataset I am using is this one:
>
>
>
http://support.sas.com/documentation/cdl/en/statug/63033/HTML/default/viewer.htm#statug_mixed_sect034.htm
>
> If I run the following code:
>
> proc mixed data=rc method=ML covtest;
> class Batch;
> model Y = Month / s;
> random Int Month / type=cs sub=Batch s;
> run;
>
> The Fixed effect coefficients match with that of R. But the random
> effect does not. Here is the R code:
>
> rc <- read.table('rc.csv', sep = ',', header=T,
na.strings=".")
>
> m1 <- lmer(formula = Y ~ Month + (Month|Batch), data = rc, REML = F)
>
> summary(m1)
>
> fixef(m1)
>
> ranef(m1)
>
> But if I change the SAS code as follows:
>
> proc mixed data=rc method=ML covtest;
> class Batch;
> model Y = Month / s;
> random Int / type=cs sub=Batch s;
> run;
>
> and the R code as follows:
>
> m2 <- lmer(formula = Y ~ Month + (1|Batch), data = rc, REML = F)
>
> summary(m2)
>
> fixef(m2)
>
> ranef(m2)
>
> both fixed and random effect coefficients match. I am unable to
> understand this discrepancy. Am I wrongly specifying the model in the
> first case?
>
> It would be helpful if someone can throw some light on this.
>
> Regards,
> Indrajit
>
> ______________________________________________
> R-help@r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
> http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>


-- 
Joshua Wiley
Ph.D. Student, Health Psychology
University of California, Los Angeles
http://joshuawiley.com/
Senior Analyst - Elkhart Group Ltd.
http://elkhartgroup.com

	[[alternative HTML version deleted]]