[R] Model building ...

[John Logsdon]

Hi

Are there any functions that de-convolute data into a given number of
clusters, rather like the NPMLE GLIM macros from Murray Aitkin and Brian
Francis?  Basically I would like to code into R the same approach but
include the possiblility of some data being censored.  In principle the
formulae are the same (just replace the likelihood function) but I haven't
managed to get my head round the model building problems.

I am looking for an elegant way of doing the following:

Consider a model where the variables are partitioned into two groups -
call them fixed and random.  I have these as data frames - each can be a
mixture in principle of vectors and factors.  Then I want to fit:

Y~fixed+F:random

where Y is the response (possibly including the censoring indicator), F is
a factor.  The response, the factor F and the random and fixed data frames
are stacked N long for N mass points. F indexes the mass point. 

I have been trying to build up the model and use gnlr() except the
complexities are beating me at the moment.  Should I use a list or a model
frame?  Can anyone advise?

John




-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._