Any variable that cannot be measured directly but is presumed to be indicated by things we can measure. Any variable presumed to be measured by responses on a multi-item self-report questionnaire and pretty much everything else of interest in what is subjective or individual is a latent variable.
Other variables can be measured directly: so-called “hard” measures such as blood chemistry, functional magnetic resonance images and sequences, height, weight and age. (But beware, “hard” measurables such as height, weight and age may have very complex meanings for us!)
Details #
The term “latent variable” is crucial to a whole raft of methods often termed “latent variable analyses” including internal reliability, factor analysis and structural relationship modelling all of which can be hugely useful in our area in which so much of interest cannot be measured directly. However, there are dangers with these methods: generically that their limitations and assumptions are often minimised and they are treated as if they distil out magical “true measurement”. As they are moderately complex and often turn on statistical and mathematical models it can be hard to critique them wisely when they are being oversold.
Try also #
- Confirmatory factor analysis
- Exploratory factor analysis
- Factor analysis
- Internal reliability/consistency
- Structural relationship modelling
Chapters #
The basic idea of latent variables runs through chapters 2 to 5 of the OMbook.
Online resources #
None currently nor likely I think.
Dates #
First created 7.i.26.
