[R] Coding for contrasts in unbalanced designs

Peter Dalgaard p.dalgaard at biostat.ku.dk
Tue Apr 3 19:20:02 CEST 2007


Guillaume Brutel wrote:
> In fact I want to test for the effects of multiple continuous and 
> categorical variables (with GLMs) which are nested in the factor levels 
> on a single continuous variable. Since the design is unbalanced, I 
> though I had to specify it in the glm formula.
>
>   
That didn't make me any wiser... What is the model and which hypotheses 
do you wish to test?
>
> Guillaume Brutel wrote:
>  > Sorry for possible cross posting but it seems that my previous 
> message was send as HTML and then scrubbed.
>  >
>  > Dear list members,
>  >
>  > I want to use a GLM with an unbalanced factor and continuous variables.
>  > My factor F has 12 unbalanced levels:
>  > 
> F=as.factor(c('A','B','C','C','C','C','D','D','D','D','D','D','E','E','E','E','E','E','F','G','G','H','I','I','J','J','J','K','L','L','L'))
>  >
>  >> summary(F)
>  >>
>  > A B C D E F G H I J K L
>  > 1 1 4 6 6 1 2 1 2 3 1 3
>  >
>  > I thus want to code contrasts appropriatly.
>  > I have looked in R documentations and R-help archives without finding 
> the information for unbalanced designs.
>  > Any help will be very useful.
>  > Thanks you very much.
>  > Guillaume Brutel.
>  >
>  >
> Er, in which sense do you mean "code contrasts"? There are some
> ambiguities relating to the use of the term "contrasts", so it isn't
> possible to tell what it is that you need. Please give an example of
> what you want to  do and explain why you think that it night come out
> wrong when data are unbalanced.
>



More information about the R-help mailing list