Longitudinal Exploratory Factor Analysis (Longitudinal EFA)
Longitudinal Exploratory Factor Analysis · Also known as: LEFA, longitudinal factor analysis, repeated-measures EFA, panel EFA
Longitudinal EFA applies exploratory factor analysis separately at each measurement occasion — or jointly across occasions — to discover whether the same latent factor structure emerges over time and whether factor loadings remain stable across waves. It is the foundational data-driven approach for examining structural change and continuity in panel and developmental research.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use longitudinal EFA when you have panel or repeated-measures item data from two or more waves and want a data-driven check of whether the same factor structure holds across time — especially early in a research programme before committing to a fixed confirmatory model. It is the right tool when the factor structure at each wave is genuinely unknown or when a previously proposed structure may not replicate across contexts or developmental periods. Do not use it as a substitute for longitudinal CFA once the structure has been established and you need formal invariance tests with fit statistics. Do not use it when you have only a single measurement occasion, when the same participants are not measured across waves, or when items differ substantially across waves, because cross-wave comparison then has no meaningful basis.
Strengths & limitations
- Data-driven: does not assume the factor structure in advance, making it appropriate for early-stage longitudinal research.
- Reveals genuine structural change across occasions rather than conflating construct change with score change.
- Tucker's congruence coefficient provides a quantitative, interpretable index of cross-wave factor similarity.
- Procrustes rotation anchors solutions to a common reference without imposing full invariance constraints, preserving exploratory flexibility.
- Flags which specific items shift in their factor salience over time, guiding item revision or theoretical interpretation.
- Does not yield formal fit statistics for the overall longitudinal model; structural comparisons remain descriptive rather than inferential.
- Factor indeterminacy across waves means that even well-matched solutions contain ambiguity about whether exactly the same factors were recovered.
- Requires large samples at each wave; instability in loadings due to small n can masquerade as genuine structural change.
- Cannot simultaneously model growth in factor means; a complementary latent growth curve or second-order longitudinal CFA is needed for that purpose.
Frequently asked
How is longitudinal EFA different from longitudinal CFA?
Longitudinal EFA is exploratory: it discovers the factor structure at each wave without specifying it in advance. Longitudinal CFA is confirmatory: you fix the structure and formally test whether it fits the data equally well at each wave. The standard workflow is to explore with EFA first, then validate and test invariance with CFA on a separate or held-out sample.
What is Tucker's congruence coefficient and what value indicates good replication?
Tucker's congruence coefficient (phi) measures the similarity between two factor loading vectors by their cosine. Values above 0.95 indicate excellent replication, values between 0.90 and 0.95 are acceptable, and values below 0.85 suggest the factors should not be considered the same construct across waves.
Do I need the same participants at every wave?
For meaningful cross-wave comparison of factor structures, yes — the same items must be administered to an overlapping or ideally identical group of participants. If the panels differ substantially (high attrition, rotating panel design), cross-wave comparison must account for possible selection effects that could reshape the observed factor structure.
Can I use longitudinal EFA with ordinal Likert items?
Yes. Use polychoric correlation matrices rather than Pearson correlation matrices as the input to EFA at each wave, and choose an extraction method compatible with polychoric input such as Principal Axis Factoring or WLSMV-based factor analysis. Ordinal EFA assumptions are then satisfied and the loadings are more accurate for Likert-type data.
What sample size do I need?
A common guideline is at least 5 participants per item, with a minimum of roughly 100–200 cases per wave for stable loadings. Smaller samples produce loadings with high sampling error, which can be misread as genuine structural change across waves.
Sources
How to cite this page
ScholarGate. (2026, June 3). Longitudinal Exploratory Factor Analysis. ScholarGate. https://scholargate.app/en/psychometrics/longitudinal-exploratory-factor-analysis
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Confirmatory factor analysisPsychometrics↔ compare
- EFAStatistics↔ compare
- Longitudinal CFAPsychometrics↔ compare
- Longitudinal Measurement InvariancePsychometrics↔ compare
- Measurement InvariancePsychometrics↔ compare
- Multilevel EFAPsychometrics↔ compare