Joost van de Weijer, Lund University Humanities Lab

Analyzing scale responses using structural equation models

Latent (unobserved) variables are those that cannot be measured directly. Examples are anxiety, agreeableness, aggression, etc. Relatively often, these variables are estimated indirectly using multi-item ordinal scale responses. A set of items related to the variable of interest are rated by participants. One way of analyzing these responses is through a structural equation model. The ratings given by the participants are used to estimate characteristics of the latent variable (e.g., the mean or the variance), and relate these to characteristics of other (observed or unobserved) variables. In the presentation, I will introduce the analysis, and illustrate it with an example from language learning.

Part 2 in a series of talks on quantitative data analysis in applied linguistics research