[R-sig-ME] glmmADMB (mixed binomal random effects)

[Ben Bolker]

On 14-02-28 05:19 PM, Julian Chen wrote:
> Hello:
> 
> I am new to this and now running a simple model of mixed effects
> negative binomial regression by using glmmADMB my data is very
> simple as follows: Counts Year Test1 Test2
> 
> 
> Here counts is the number of students annually admitted (could be
> zero) Year 1 to 6 (totally 6 numbers represnts six years) Test 1 is
> the score of test one range from (0 to 100) Test 2 is the score of
> test two range from (0 to 100)
> 
> Here I asssume Variable Year is random effects. The following is
> the model I used: rnb1<-glmmadmb(ACC ~ F1 + F2 + (1|YR), data=dat,
> family="nbinom", link="log");
> 
> But I always got the wrong message as below:
> 
> rnb1<-glmmadmb(Counts ~ Test1 + Test2 + (1|Year), data=dat,
> family="nbinom", link="log"); Error in Droplevels(eval(parse(text =
> x), data)) : all grouping variables in random effects must be
> factors
> 
> I have no idea what I should do now? Please give me some comments.
> Many thanks!
> 
  [cc'ing to r-sig-mixed-models]

  make sure that your grouping variable (Year) is a factor.

  Year is coded as numeric.  Your choices:

1. convert to a factor on the fly:
    Counts ~ Test1 + Test2 + (1|factor(Year)
2. convert to a factor in the data frame:
   dat <- transform(dat,Year=factor(Year))
3. make a new factor variable in the data frame:
   dat <- transform(dat,fYear=factor(Year))
   Counts ~ Test1 + Test2 + (1|fYear)

#1 is quickest but might be slightly less robust sometimes
#2 is robust but you might want to keep Year as numeric for other purposes
#3 keeps year as numeric and adds an additional factor version of the
variable