[R] Problems with Z in rhierMnlRwMixture using bayesm

[ea819]

Dear All,
I am using rhierMnlRwMixture in the bayesm package for the analysis of data
from a choice experiment. I am trying to follow the margarine example set
out in the bayesm manual (p.28). However, after several attempts I keep
getting an error message with regards to my Z matrix as below. 
> Error in Z %*% t(matrix(olddelta, ncol = nz)) : 
>requires numeric/complex matrix/vector arguments

I think the problem is arising from the following lines in the code that I
am running. 
>Z=NULL
>nlgt=length(lgtdata)
>for(i in 1:nlgt) {
>Z=rbind(Z,Demog[Demog[,1]==lgtdata[[i]]$id,3:9])

I am creating a Z matrix from the variables included in the “Demog”
spreadsheet columns 3 to 9 which includes the demographic variables of the
respondents. Then the rbind matches these columns with the ids in the
lgtdata (which contains the choice attributes). I am not sure if I am
interpreting this code correctly or if I need to amend it. Can anyone help
me with the interpretation of the below and whether I would need to specify
it differently? 

Also below is the full code that I am running.
select= c(1,2,3)  
 id=levels(as.factor(ChoiceAtt[,1]))
 lgtdata=NULL
 nlgt=length(id)
 p=length(select)  
 ind=1
 for (i in 1:nlgt) {
    nobs=sum(ChoiceAtt[,1]==id[i])
       data=ChoiceAtt
       y=data[,2]
       names(y)=NULL
 	  X=createX(p=p,na=1,Xa=data[,3:5],nd=NULL,Xd=NULL,INT=TRUE,base=1)
        lgtdata[[ind]]=list(y=y,X=X,id=id[i]); ind=ind+1
     }
 nlgt=length(lgtdata)

Z=NULL
nlgt=length(lgtdata)
for(i in 1:nlgt) {
Z=rbind(Z,Demog[Demog[,1]==lgtdata[[i]]$id,3:9])
 
}
Z=log(Z)
 Z[,1]=Z[,1]-mean(Z[,1])
 Z[,2]=Z[,2]-mean(Z[,2])
 
 keep=5
 R=5000
 mcmc1=list(keep=keep,R=R)

out=rhierMnlRwMixture(Data=list(p=p,lgtdata=lgtdata,Z=Z),Prior=list(ncomp=1),Mcmc=mcmc1)

Many thanks for all your help and time. 

Regards, 
Elcin


--
View this message in context: http://r.789695.n4.nabble.com/Problems-with-Z-in-rhierMnlRwMixture-using-bayesm-tp3717887p3717887.html
Sent from the R help mailing list archive at Nabble.com.