[R] R lmer & SAS glimmix

[Bert Gunter]

It's due to the different coding of contrasts. Post to R-sig-mixed
models for a fuller explanation (though it has nothing to do with
mixed models; the same would happen with lm) or search on a suitable
phrase (e.g. contrast coding in SAS versus R) in R archives or via
google etc.

-- Bert

On Mon, Nov 12, 2012 at 6:40 AM, Sophie GUERIN <s.guerin at has-sante.fr> wrote:
> Hi,
>
> I am trying to fit a model with lmer in R and proc glimmix in SAS. I have
> simplified my code but I am surprised to see I get different results from
> the two softwares.
>
> My R code is :
> lmer(y~age_cat + (1|cat),data=fic,family=binomial(link = "logit"), NaGQ=1)
>
> My SAS code is :
> ods output Glimmix.Glimmix.ParameterEstimates=t_estimates;
> proc glimmix data=tab_psi method=laplace;
> class age_cat cat;
> model psi (event='1') = age_cat / solution dist=B link=logit ;
> random  intercept / subject=cat;
> run;
>
> >From R, I get the following fixed effects
> (Intercept) age_cat2. 76-85 ans age_cat3. 66-75 ans age_cat4. 41-65 ans
>          -3.5766898          -0.0159466          -0.1919500 -0.4834741
> age_cat5. 18-40 ans
>          -1.2843977
>
> But from SAS I get :
>                                Valeur      Erreur              Valeur
>          Effet        age_cat          estimée        type      DDL    du
> test t    Pr > |t|
>
>          Intercept                     -4.8608      0.2859        3 -17.00
>      0.0004
>          age_cat      1. >85-108 a      1.2841      0.2589    168E3  4.96
>    <.0001
>          age_cat      2. 76-85 ans      1.2681      0.2528    168E3  5.02
>    <.0001
>          age_cat      3. 66-75 ans      1.0921      0.2529    168E3  4.32
>    <.0001
>          age_cat      4. 41-65 ans      0.8006      0.2535    168E3  3.16
>    0.0016
>          age_cat      5. 18-40 ans           0           .        .   .
> .
>
> Even the intercept is different, but I can't find why. Has anyone an idea?
>
> Thanks in advance,
>
> Sophie
>         [[alternative HTML version deleted]]
>
>
> ______________________________________________
> R-help at 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.
>



-- 

Bert Gunter
Genentech Nonclinical Biostatistics

Internal Contact Info:
Phone: 467-7374
Website:
http://pharmadevelopment.roche.com/index/pdb/pdb-functional-groups/pdb-biostatistics/pdb-ncb-home.htm