[Forum SIS] DEI - Università di Catania: AVVISO DI SEMINARI

[Roberto Di Mari]

********************************************************************************
AVVISO DI SEMINARIO
********************************************************************************

Il Dipartimento di Economia e Impresa dell’Università di Catania ha
organizzato il seminario:



*A two-step approach to Latent Class Analysis for models with complex
dependencies*

*Zsuzsa Bakk*

*Institute of Psychology, Methodology & Statistics Unit*

*Leiden University*







Mercoledì 5 Aprile dalle ore 10:00

In Aula C, Palazzo Fortuna



Università di Catania

Palazzo delle Scienze, Corso Italia, 55, Catania





*Abstract*:



Latent class analysis is a statistical method widely used by social and
behavioral scientists for building typologies and classifications based on
a set of observed characteristics. Examples include typologies of
individual’s attitudes based on survey questions, subtypes of schizophrenia
patients derived from recorded mood symptoms, and classifications of
consumers inferred from stated or revealed preferences.

Researchers usually relate the classifications to other variables, for
example to assess how attitudes vary by education or nationality. This
analysis is usually performed using the method of three-step latent class
analysis. While this approach works well in most situations, it fails when
the way the latent classes are related to the observed responses differs
across subgroups. This often happens for example in cross-national surveys,
where the survey questions have a different meaning in some countries due
to differences in translation. Using the three-step approach it is
impossible to identify and account for these types of differences, leading
often to meaningless or unfair conclusions.

I propose a novel two-step latent class approach that can easily identify
and adequately model subgroup differences, thus avoiding erroneous
conclusions. Furthermore the approach is general enough to be applicable in
any situation that requires stepwise latent class modelling.


I colleghi interessati sono cordialmente invitati a partecipare.

Roberto Di Mari
-------------- parte successiva --------------
Un allegato HTML è stato rimosso...
URL: <http://www.stat.unipg.it/pipermail/sis/attachments/20170324/f20e4b8e/attachment.html>