[R] lme function problem

[Marc Schwartz]

On Aug 23, 2009, at 10:14 AM, Brittany Hall wrote:

> Hello,
>
> I am fairly new to R and having a problem with the lme command. I  
> have searched on forums, read the Fox 2002 chapter, and R help, but  
> the suggestions that I have tried have not helped me.
>
> My data file is called Acsdata.
>
> This is my script:
>
> Acsdata.1 <- lme(Acsdata$gsi ~ Acsdata$asitotal + Acsdata$famstrto +  
> as.factor(Acsdata$GEN) + Acsdata$asxfs + Acsdata$asxpar + Acsdata 
> $asxgp + Acsdata$asxfsxpa + Acsdata$asxfsxgp, random = ~1|Acsdata 
> $family)
>
> I keep getting the error: object "gsi" not found. However, when I  
> use the summary function, R is able to recognize gsi.
>
> I welcome any suggestions.
>
> Thanks

Be sure that the Acsdata data frame is not attach()ed. That can cause  
certain conflicts.

Then, be sure to utilize the 'data' argument that is available in R  
functions that take formulae as arguments. So use:

Acsdata.1 <- lme(gsi ~ asitotal + famstrto + as.factor(GEN) + asxfs +  
asxpar + asxgp + asxfsxpa + asxfsxgp,
                  random = ~1|family, data = Acsdata)


I also think that it would be 'cleaner' to coerce 'GEN' to a factor  
before calling lme(). It results in more readable output from the  
function and it's methods.

See 'An Introduction to R', available in your R distribution or online  
at http://cran.r-project.org/manuals.html for general information on  
using modeling functions.

Addtionally, as you progress through this, you may have follow on  
questions and there is a focused R e-mail list on mixed models. See https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models 
  for more information.

Finally, in addition to John's excellent book that you reference  
above, the seminal reference for lme() is:

Jose C. Pinheiro and Douglas M. Bates
Mixed-Effects Models in S and S-Plus
Springer, 2000. ISBN 0-387-98957-0


HTH,

Marc Schwartz