Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

Re: st: Latent class analysis

From   Stas Kolenikov <[email protected]>
To   [email protected]
Subject   Re: st: Latent class analysis
Date   Thu, 28 Oct 2004 13:21:03 -0400

-gllamm- can handle two classes with the -np(f)- option (or whatever
is the one that requests the non-parametrics specfication of the
likelihood). Of course it cannot perform any real LCA.

I was looking into this a few weeks ago, and figured out I would have
to code this in -ml- if I need to. It is not too difficult, it just
requires some good bookkeeping.

Actually, there may be a problem with the way MLE is implemented in
Stata. If you hit the boundary (as you may well with probabilities in
your LCA at 0 or 1), and code the likelihood as . (missing) for the
implausible values, Stata's -ml- maximizer will be unhappy about it,
and report "cannot compute likelihood -- missing values encountered".
My question to Stata Corp.: is there any way around this? Parameter
transformation wouldn't work, as I do need to include 0's and 1's as
valid parameter values, and my parameter set is closed interval [0,1].

On Thu, 28 Oct 2004 13:17:02 +0100, Modesto G Gayo-Cal
<[email protected]> wrote:
> I've been trying to find out whether it is possible to do a latent class
> analysis using STATA. I found the GLAMM programme and I
> skimmed the manual, but I didn't find any proper answer. It seems
> that it is been designed to do more sophisticated analysis, and
> maybe it's been taked for granted that the potential users would
> know how to deal with a simple latent class analysis. This is not
> the case. Therefore, could someone give some answer to the next
> logically linked questions?

Stas Kolenikov
*   For searches and help try:

© Copyright 1996–2024 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index