Description
A person-specific factor model that lets each individual's indicators load differently while holding the latent dynamics equal across people, so the same process is compared fairly.
The idiographic filter starts from the observation that one construct can surface through different indicators in different people: stress as stomach trouble in one child, tremor in another. Instead of demanding identical loadings (classic factorial invariance), it splits the data box into one intensive time series (P-technique dataset) per person and fits a joint state-space model in which loadings are free per person but latent auto-regressions, cross-regressions and factor correlations are constrained equal. Good fit means the same latent process is being measured even though it shows itself differently; if loadings turn out identical, the model collapses to ordinary invariance. Measurement equivalence is thereby redefined dynamically: same process, not same trait expression. iFACE extends the approach to twin pairs to estimate person-specific heritability. Its rigour lies in an explicit, testable equality constraint at the latent level.
Attributes
- Discipline
- Survey & Psychometrics
- Loss type
- Unmigrated
- Stage
- Evaluation, Post-launch
- Data
- Quantitative
- Effort
- Months
Steps
- Collect multivariate time series on the same variables for each of several people, with enough occasions per person for P-technique factor analysis (typically 50 or more).
- Specify a state-space model with person-specific factor loadings and a latent process model (auto- and cross-regressions, factor correlations) constrained equal across persons.
- Fit the joint model and test it against a model with loadings also constrained equal; if loadings do not differ, conclude ordinary factorial invariance.
- If the idiographic model fits well, interpret each person's loading pattern alongside the shared latent dynamics.
- Derive second-order factors where needed and express them on the observed variables by Schmid-Leiman or Cattell-White transformation.
- Report which indicators carry the construct for whom, and use the shared latent scores for between-person comparison.
Sources
- Handbook of Child Psychology Vol 1 (Lerner et al., 2015)
Also known as
- Idiographic filter
- IF
- iFACE
- Dynamic measurement equivalence
- Idiographic filter (IF) and iFACE
Categories
- Analytics & Behaviour
- Survey & Psychometrics