[R-sig-ME] 'fixed' covariates in nlmer

[Laurent Gentzbittel]

Dear all,


I need to fit a logistic curve to disease index for plants subjected to
a pathogen. I have longitudinal data (the disease is monitored on
different plants as a function of time) and I would like to use lme4 and
nlmer function.

fit1 <- nlmer( score ~ SSlogis(dpi, Asym, xmid, scal) ~ (Asym + xmid |
plant),
          start=c(Asym=3.9 ,xmid= 12 ,scal= 1.4)
          data=mydata)

do the job by fitting random 'plant' effects for the Asym and xmid
parameters and also estimate population values (fixed effects) for Asym,
xmid and scal

However, not all the plants have the same genotype and I see on the
curves that some genotypes have different Asym and xmid values. Thus I
need to include a fixed 'genotype' effect for those parameters. I was
not able to find any line of code or example, nor in the lme4
documentation, that could help me to specify a 'genotype' fixed effect
for some of the parameters

I tried :
fit2 <- nlmer( score ~ SSlogis(dpi, Asym, xmid, scal) ~  (Asym|genotype)
+ (Asym + xmid | plant))

that fit a RANDOM genotype effect for Asym BUT not a FIXED effect as I
expected.

I also tried (mimicking the lmer syntax) :

fit3 <- nlmer( score ~ SSlogis(dpi, Asym, xmid, scal) ~  genotype +
(Asym + xmid | plant))

yet that syntax does not add a fixed 'genotype' effect for each three
parameters


How do I must specify these fixed effects ?

ps : I posted a similar mail one year ago, did not get answers and
survived using the lme package instead of the lme4 one. However, I
really would like to change.

Thanks for your help.
Laurent

-- 
Prof. L. GENTZBITTEL            Phone: +33 (0)5 62 19 35 96
INP - Agro Toulouse             Fax  : +33 (0)5 62 19 35 89
Symb. & Path. Plantes   - IFR40
18, Chemin de Borde Rouge - Auzeville Tolosane
31326 CASTANET TOLOSAN - FRANCE
E-mail:                         gentz at ensat.fr
Plant genomes synteny  :       
http://bioinfo.genopole-toulouse.prd.fr/iccare/
Linux user #301231