A specific set of factor analytic models these days always conducted using a confirmatory factor analysis program.
Details #
These are typically used where we may have item data from one or more questionnaires, as ever, completed by individuals independent of each other. Here the model that is tested for fit to the data assumes that items load on one “first order factors” as in an ordinary exploratory or single order confirmatory factor analysis. However, hierarchical models add one or even more “second order factors” to the model.
As in non-hierarchical models the idea is that the first order factors reflect specific interpersonal differences, e.g. the model might be that the 12 Somatization (SOM) items of the SCL-90-R might share variance coming from some differences beteween people in how much they somatise and that the 10 items of the ObsessionCompulsion (O-C) scale might likewise reflect differences between people in their tendency to obsessional compulsive problems and the 13 Depression (DEP) items reflect differences in depression problems and it might be that these first order factors, SOM, O-C and DEP are correlated. What a hierarchical model adds is the idea that the correlation between first order factors results from a more general factor of our proneness to have any of these problems.
This has some similarities with a bifactor model.
Try also #
- Bifactor models
- Confirmatory factor analysis
- Exploratory factor analysis
- Factor analysis
- Network analysis
- Psychometrics
Chapters #
Not covered in the OMbook.
Online resources #
As for bifactor models and perhaps the whole area of factor analyses and similar methods this probably needs an Rblog post with plots to explain it better: coming but don’t hold your breath!
Dates #
First created 20.i.25.
