Measurement: Measurement Models and their Applications
Content
Measurement is one of the central elements of empirical research, intimately linked to observation, conceptualisation, and validity. Moreover, a variety of quantitative methods derive new measurements from existing ones.
This clinic covers the following topics:
Measurement theory
What are the characteristics of what we call measurement (including categorisation and typology construction), and how does this relate to observation, conceptualisation and validity?
Model-based measurement
Measuring variables we are interested in using multiple indicators is preferable to doing so with single indicators in terms of reliability, precision and construct validity. Measurement with multiple items is exceedingly common in a variety of social research areas. Social and economic indicators about localities, regions or countries (e.g., literacy, life-expectancy, GDP per capita, internet penetration, mobility, and so on) are often assumed to reflect a small number of ‘underlying’ characteristics of those geographical units. In survey research, answers to a variety of questions are often thought of as reflecting a smaller number of attitudes or orientations. In educational and psychological testing, a large number of test-items are assumed to all express a single underlying trait, etc.
In such cases, the question is how we are to know whether or not a set of items measures the same phenomenon? And subsequently, how to use this knowledge in getting better measurement. Measurement models help us to answer this question.
Factor analysis
One of the most widely know models for analysing multiple indicators is factor analysis. We will introduce the factor analysis (FA) model and its relation to principal component analysis (PCA). We will distinguish between exploratory factor analysis (EFA) and confirmatory factor analysis (CFA). CFA is also known as the latent variable measurement aspect of structural equation modelling (LISREL, AMOS, etc.).
FA applicability issues
Although FA is often touted as the method par excellence for the measurement of latent variables from multiple indicators, it is not suited in a number of circumstances and, when applied in an inappropriate context can yield incorrect and misleading results. The clinic diagnoses the conditions under which FA is, or is not suitable, and provides a first introduction to alternative methods for model-based multiple-item measurement. These models will be treated in more detail in the
Scaling clinic.
Latent Class Analysis (LCA)
LCA is a model-driven approach to nominal-level latent variables. It is a modern successor to traditional segmentation analysis.
This clinic will be supplemented with online learning materials which can be accessed after the event.
Prerequisites
Introductory understanding of statistics, e.g. the mean, standard deviation and standard error of a variable; correlation and covariance. If you are attending the clinic as part of our Researcher Development Initiative then you automatically fulfil the prerequisites.