[ADMB Users] crappy software

[dave fournier]

    I built the following r script to test glmer.
   I modifed the simulator to produce an R input file, r.dat.

   the script is

    library("lme4")
    system("./simulator",intern=TRUE)
    dat=read.table("r.dat",header=TRUE)
    fm <- glmer(Y~ 1 + x + (1 + x|id), data=dat,family=binomial)
    sink("r.out")
    fm
    sink()
    system("grep \"^x   \"    r.out >> xvalues ",intern=TRUE)

It outputs the estimate for b1 and its std dev into the file xvalues
The difference is that the sign of b1 is changed  (using 1/(1+exp(+-x)) )

I only did 60 simulations to date just to see how it works. The type 1 error
appears to be around 30%.  Of course I don't know anything about R
(thank God) so maybe the formula is wrong.

       fm <- glmer(Y~ 1 + x + (1 + x|id), data=dat,family=binomial)







-------------- next part --------------
An embedded and charset-unspecified text was scrubbed...
Name: simulator.tpl
URL: <http://lists.admb-project.org/pipermail/users/attachments/20111031/5d45f74e/attachment.ksh>