|
Title
|
Room
|
|
A Graphical Introduction to Latent Class Analysis
|
M-International Salon E
|
|
Date / Time
|
Sponsor
|
Type
|
|
08/06/2001
1:00 PM
-
5:00 PM
|
ASA, Section on Statistical Graphics*
|
Other
|
|
Organizer:
|
n/a
|
|
Chair:
|
n/a
|
|
Discussant:
|
|
|
CE Presenter
|
Jay Magidson
Jay Magidson
Jeremy Magland
Jeremy Magland
|
Description
Use of latent class (LC) and finite mixture models is growing rapidly because of 1) major developments in maximum likelihood computational algorithms for these models and 2) the lack of restrictive assumptions underlying the general model (Vermunt and Magidson, 2000a). In this short course we introduce LC as a probability model and focus on 2 important special cases - that of cluster analysis and factor analysis for combinations of nominal, ordinal, and continuous variables. Graphical displays of results will be emphasized.
The close relationship between LC and correspondence analysis (Van der Heijden et.al. 1999; Magidson and Vermunt, 2000) has paved the way for new intuitive graphical approaches for displaying the results of LC models as an alternative to the traditional output (parameter estimates, standard errors, p-values, etc.). Both the traditional and graphical kinds of output are implemented in a new computer program called Latent GOLD (LG), which will be used for demonstrations (Vermunt and Magidson, 2000b). By using examples of real data and focusing on the graphical output, prerequisites for attendees can be kept to a minimum. The only requirement is familiarity with cross-tabulation of categorical data and chi-squared test for independence.
|
|
