[R] multinomial conditional logit models

[Ott Toomet]

Hi,

 | I've been trying to implement an MCL model in R and have been mostly
 | succesful. However I'm still stuck on a few points and I'm hoping
 | someone can point out how to do fix them.
 | 
 | To estimate an MCL model, the data must be restructured into a
 | person-choice file. 
 | 
 | * Each case must be duplicated "ncat" times ("ncat" is the number of
 | categories of the dependent variable)
 | * Each case must be identified by a strata variable (id)
 | * Each duplicate must be identified by a variable indexing the
 | choices (newy)
 | * A dichotomous variable must indicate which duplicate corresponds
 | with the respondent's choice (didep)
 | 
 | I've done this as follows:
 | 
 | mclgen <- function (datamat, catvar) {
 | 	ncat <- nlevels(catvar)
 | 	id <- 1:length(catvar)
 | 	datamat <- cbind(id,datamat)
 | 	perschoice <-NULL
 | 	for (i in 1:ncat) {
 | 		perschoice<-rbind(perschoice,datamat)
 | 	}
 | 	perschoice <- perschoice[sort.list(perschoice$id),]
 | 	perschoice$newy <- gl(ncat, 1, length=nrow(perschoice))
 | 	dep <- parse(text=paste("perschoice$", substitute(catvar), sep=""))
 | 	perschoice$depvar <- as.numeric(eval(dep) == perschoice$newy)
 | 	perschoice
 | }
 | 
 | This works but I wonder if it's the most efficient method. I tried
 | using "rep" rather than "rbind" in a loop but that replicated the
 | dataset horizontally, not vertically. Is this the best solution?

Well, in general it is not efficient to add something to a matrix in a
cycle.  You should in either make the datamat in required size and
fill the required parts of it, or in this case you may consider e.g.

> datamat <- matrix(1:4, 2, 2)
> matrix(t(datamat), 4, 2, byrow=T)
     [,1] [,2]
[1,]    1    3
[2,]    2    4
[3,]    1    3
[4,]    2    4

 | I also finally figure out how to include the argument "catvar" in a
 | transformation involving another dataset but this solution seems very
 | complicated (eval+parse+substitute). Is there a simpler solution?

Unfortunately, I was not able to understand quite what are you doing
here.  Where is the dataset logan?  I did not find it on
http://www.xs4all.nl/~jhckx.  If I could get a simple example running,
I can try further.

best wishes,

Ott