T.S. Lim <[EMAIL PROTECTED]> wrote:
> In the software package Latent GOLD
>
> http://www.statisticalinnovations.com/lg/highlights.htm
>
> you can do regression (continuous or discrete response variable) using
> mixture modeling. I'm interested in the statistical model behind the
> method. Anyone knows of (online) papers that discuss it? Thanks.
Well, an obvious way to implement a regression using a mixture model
is to construct a mixture density over all the variables (predictors
and response) and then compute the conditional density of the response
given the predictors. The latent variable in this kind of scheme is
the indicator which tells to which bump each case is assigned.
However, that doesn't appear to be the way that Latent GOLD (gotta
love these trade names :) goes about it. It seems they work directly
with a mixture of conditional densities without going through the
intermediate step of constructing the joint density. Here, see for
yourself: http://www.statisticalinnovations.com/lg/lg_app3.pdf
This paper contains a number of references that may be helpful.
For what it's worth,
Robert Dodier
--
``Socrates used to meditate all day in the snow, but Descartes'
mind worked only when he was warm.'' -- Bertrand Russell
Sent via Deja.com
http://www.deja.com/
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
