STATISTICAL COMPUTING SECTION, ROYAL STATISTICAL SOCIETY.

Half day meeting on
Latent class models and finite mixture models

Wednesday, 27th November. 14.00 - 17.15
Royal Statistical Society, Errol Street, London

Classical latent class analysis is a technique for probabalistic
clustering, finding subtypes of related cases from multivariate categorical
data. This half day meeting discusses extensions and developments in latent
class modelling. Software implementations in Latent GOLD and STATA will be
discussed, with applications from medicine, management and the social sciences.

The meeting will be held at the Royal Statistical Society. There is no
charge, and no prior registration is needed. See
http://www.rss.org.uk/about/direction.html for details on how to find the RSS.

14.00 Applications of Latent Class Analysis: An Introduction to the
Technique and the LatentGOLD Software.
Jeroen K. Vermunt Department of Methodology and Statistics
Tilburg University

The basic idea underlying latent class and finite mixture models is
extremely simple: one or more of the parameters of a postulated statistical
model differ across unobserved subgroups. This idea has several seemingly
unrelated applications, such as clustering, scaling, density estimation,
and random-coefficients modeling. The most important types of applications
will be introduced using empirical examples from sociology, medicine,
psychology and marketing. The LatentGOLD program will also be described,
developed by the speaker in collaboration with Jay Magidson.

14.50 Using the latent class model for the analysis of misclassified data
Ardo van den Hout and Peter van der Heijden, Department of Methodology and
Statistics, Utrecht University

In randomized response experiments, statistical disclosure control and
scales with a known specificity and sensitivity data are misclassified with
a known misclassification scheme. We show how the latent class model can be
used to analyse such data.

15. 45 Tea/coffee break

16.15 Latent classes: discrete random effects and factors in GLLAMM
Andrew Pickles1, Sophia Rabe-Hesketh2 and Anders Skrondal2,
School of Epidemiology and Health Science,
1University of Manchester 2Institute of Psychiatry,London

The GLLAMM (Generalized Linear Latent and Mixed Models) procedure that is
implemented in Stata is able to fit random effects and latent variables
with distributions that can be assumed normal, discrete classes or
unspecified (estimated using NPML). We explore the range of latent class
models that can be cast in this framework, discuss some identification
issues, and describe some examples.

17.15 Close of meeting