> I use kinship package of R and it doesn't create a kinship matrix with
repeated id.
The kinship package is out of date, use the kinship2 and coxme packages (it was
split into
two parts). Then the problem you describe no longer exists -- the kinship
function no
longer requires each subject id to be unique across all pedigrees.
pedlist <- with(mydata, pedgiree(id, fa_id, mo_id, sex, famid=famid))
kmat <- kinship(pedlist)
fit <- coxme(Surv(time, status) ~ x1 + x2 +... + (1| famid/id), varlist=kmat)
The pedlist object contains all the families, plot(pedlist[4]) would pull out
and plot the
4th family.
The column labels of kmat will be of the form "famid/id"
To solve your original problem (though you don't need to) create a single
unified id.
uid <- paste(famid, id, sep='/')
idlist <- unique(uid)
newid <- match(uid, idlist)
newmom <- match(paste(famid, mo_id, sep='/'), idlist)
newdad <- match(paste(famid, fa_id, sep='/'), idlist)
Terry Therneau
author of kinship and coxme, but not :-) the maintainer of kinship2