Institute for Mental and Organisational Health, FHNW
05 June, 2025
R?At the end of this course, you…
R.I try to cite relevant literature throughout the workshop. Large parts of the theoretical introductions are based on the following sources:
How many constructs underly the following measured variables?
| V1 | V2 | V3 | V4 | V5 | V6 | |
|---|---|---|---|---|---|---|
| V1 | 1.00 | |||||
| V2 | .90 | 1.00 | ||||
| V3 | .90 | .90 | 1.00 | |||
| V4 | .90 | .90 | .90 | 1.00 | ||
| V5 | .90 | .90 | .90 | .90 | 1.00 | |
| V6 | .90 | .90 | .90 | .90 | .90 | 1.00 |
How many constructs underly the following measured variables?
| V1 | V2 | V3 | V4 | V5 | V6 | |
|---|---|---|---|---|---|---|
| V1 | 1.00 | |||||
| V2 | .90 | 1.00 | ||||
| V3 | .90 | .90 | 1.00 | |||
| V4 | .00 | .00 | .00 | 1.00 | ||
| V5 | .00 | .00 | .00 | .90 | 1.00 | |
| V6 | .00 | .00 | .00 | .90 | .90 | 1.00 |
How many constructs underly the following measured variables?
| V1 | V2 | V3 | V4 | V5 | V6 | |
|---|---|---|---|---|---|---|
| V1 | 1.00 | |||||
| V2 | .00 | 1.00 | ||||
| V3 | .00 | .00 | 1.00 | |||
| V4 | .00 | .00 | .00 | 1.00 | ||
| V5 | .00 | .00 | .00 | .00 | 1.00 | |
| V6 | .00 | .00 | .00 | .00 | .00 | 1.00 |
From the course description:
Factor analysis is a multivariate statistical method wherein we strive to uncover a structure or patterns in the associations between variables (e.g., items of a questionnaire) and represent them more parsimoniously with a smaller set of underlying latent variables, called factors. These factors are thought to constitute unobservable, internal attributes, that influence or cause the way observable, i.e., manifest behavior is expressed.
\(\rightarrow\) Parsimoniously represent the structure of correlations between measured variables (MVs; e.g., questionnaire, aptitude test, etc.) to a smaller set of common factors (CFs; latent variables).
Factor analysis is based on the common factor model.
\[ P=\Lambda\Phi\Lambda^T+\Theta_\delta \]
| MV1 | MV2 | MV3 | MV4 | MV5 | MV6 | |
|---|---|---|---|---|---|---|
| MV1 | 1.00 | |||||
| MV2 | \(\rho_{2,1}\) | 1.00 | ||||
| MV3 | \(\rho_{3,1}\) | \(\rho_{3,2}\) | 1.00 | |||
| MV4 | \(\rho_{4,1}\) | \(\rho_{4,2}\) | \(\rho_{4,3}\) | 1.00 | ||
| MV5 | \(\rho_{5,1}\) | \(\rho_{5,2}\) | \(\rho_{5,3}\) | \(\rho_{5,4}\) | 1.00 | |
| MV6 | \(\rho_{6,1}\) | \(\rho_{6,2}\) | \(\rho_{6,3}\) | \(\rho_{6,4}\) | \(\rho_{6,5}\) | 1.00 |
| CF1 | CF2 | |
|---|---|---|
| MV1 | \(\lambda_{1,1}\) | \(\lambda_{1,2}\) |
| MV2 | \(\lambda_{2,1}\) | \(\lambda_{2,2}\) |
| MV3 | \(\lambda_{3,1}\) | \(\lambda_{3,2}\) |
| MV4 | \(\lambda_{4,1}\) | \(\lambda_{4,2}\) |
| MV5 | \(\lambda_{5,1}\) | \(\lambda_{5,2}\) |
| MV6 | \(\lambda_{6,1}\) | \(\lambda_{6,2}\) |
| MV1 | MV2 | MV3 | MV4 | MV5 | MV6 | |
|---|---|---|---|---|---|---|
| CF1 | \(\lambda_{1,1}\) | \(\lambda_{2,1}\) | \(\lambda_{3,1}\) | \(\lambda_{4,1}\) | \(\lambda_{5,1}\) | \(\lambda_{6,1}\) |
| CF2 | \(\lambda_{1,2}\) | \(\lambda_{2,2}\) | \(\lambda_{3,2}\) | \(\lambda_{4,2}\) | \(\lambda_{5,2}\) | \(\lambda_{6,2}\) |
| CF1 | CF2 | |
|---|---|---|
| MV1 | 1.00 | |
| MV2 | \(\phi_{2,1}\) | 1.00 |
\(\rightarrow\) When CFs are orthogonal, \(\Phi\) is an identity matrix.
| U1 | U2 | U3 | U4 | U5 | U6 | |
|---|---|---|---|---|---|---|
| U1 | \(\delta_{ 1,1}\) | |||||
| U2 | 0 | \(\delta_{2,2}\) | ||||
| U3 | 0 | 0 | \(\delta_{3,3}\) | |||
| U4 | 0 | 0 | 0 | \(\delta_{4,4}\) | ||
| U5 | 0 | 0 | 0 | 0 | \(\delta_{5,5}\) | |
| U6 | 0 | 0 | 0 | 0 | 0 | \(\delta_{6,6}\) |

\[ P=\Lambda\Phi\Lambda^T+\Theta_\delta \]
Assuming orthogonal factors:
Assuming correlated factors:
We look at two kinds of factor analysis:
R-packages: {psych}, {lavaan}, {EFAtools}R-package: {lavaan}
Introduction to Factor Analysis - GSP 2025 - University of Basel