AQMeN - Sensitivity of conclusions to incorrect assumptions about cross-national measurement equival (few places remaining)
Date:
28/05/2012
Organised by:
Applied Quantitative Methods Network (AQMeN)
Presenter:
A seminar by Dr Jouni Kuha, Jouni Kuha, Irini Moustaki and Sally Stares, London School of Economics and Political Science.
Level:
Entry (no or almost no prior knowledge)
Contact:
Description:
A seminar by Dr Jouni Kuha, Jouni Kuha, Irini Moustaki and Sally Stares, London School of Economics and Political Science.A seminar by Dr Jouni Kuha, Jouni Kuha, Irini Moustaki and Sally Stares, London School of Economics and Political Science.
This lecture is open to all AQMeN members and complements the Latent Trait & Latent Class Analysis training workshop.
Abstract
In modelling of cross-national survey data using latent variable models, substantive interest usually focuses on the distributions of the latent variables, for example on cross-national comparisons of averages or on models for associations between the latent variables. In contrast, the measurement models for the observed indicator variables are typically of lesser interest, and any complexities in these models a source more of nuisance than of excitement.
Nevertheless, it is necessary to consider the problem of measurement in some detail before comparisons between groups can be made. One potentially serious problem in cross-national analyses is the possibility of that the survey measures do not function identically across countries, i.e. that the measurement may lack cross-national equivalence. If non-equivalence of measurement is present but is ignored in the analysis, this can invalidate cross-national comparisons of latent constructs - at least in principle. In this talk, we present a sensitivity analysis of when ignoring lack of measurement equivalence does or does not seriously distort substantive conclusions, with the aim of providing researchers with some guidelines for cross-national comparisons. The focus is on latent class and latent trait models, which are used to analyse categorical survey questions. The results are also illustrated with examples from analyses of cross-national surveys.
Venue: University of Edinburgh, (exact location TBC)
Cost:
n/a
Website and registration:
http://aqmen.ac.uk/events/latent
Region:
Scotland
Keywords:
Survey and Questionnaire Design, Small Area Estimation, Latent Variable Models
Related publications and presentations:
Survey and Questionnaire Design
Small Area Estimation
Latent Variable Models