Description
Models each person's own intensive repeated measurements, with lagged effects, and generalises across people only where person-specific structures agree.
Molenaar showed through ergodic theory that results from between-person analyses apply to individuals only if the process is stationary and identical across people, conditions that learning and development violate. Analysis therefore starts within the person. P-technique factor-analyses one person's occasions-by-variables matrix; dynamic factor analysis adds lagged loadings or latent auto- and cross-regressions in state-space form to capture carry-over from one occasion to the next. GIMME builds on this: it fits a unified SEM of contemporaneous and lagged paths to each person's series, adds a path to the group structure only if it significantly improves fit for at least 75% of individuals, then searches person-specific paths. In a simulation, pooling three nearly identical families produced paths that existed in none of them. Typical data are P-data: few people, many occasions (often over 50).
Attributes
- Discipline
- Statistics & Modelling
- Loss type
- Unmigrated
- Stage
- Discovery, Post-launch
- Data
- Quantitative
- Effort
- Months
Steps
- Decide whether the question concerns between-person differences or within-person processes, using Cattell's data box (persons x variables x occasions).
- Collect intensive repeated measures of the same variables from each person at many, ideally equally spaced, occasions (50 or more via experience sampling, diaries or telemetry).
- Fit a P-technique or dynamic factor (state-space) model per person, including lagged effects; allow time-varying parameters if the process is nonstationary.
- For network questions run GIMME: free group-level paths one at a time, keeping each only if significant for at least 75% of people, and prune paths that fall below that majority.
- Use the group structure as each person's starting model, add person-specific paths while they improve fit, prune non-significant ones, and fit a final confirmatory model per person.
- Compare person-specific structures and aggregate only where homogeneity holds.
Sources
- Handbook of Child Psychology Vol 1 (Lerner et al., 2015)
- Designing an Experience Sampling Study (Dejonckheere & Erbas, 2021)
- A Beginner's Guide to Structural Equation Modeling (Schumacker & Lomax, 2004)
- The SAGE Handbook of Quantitative Methods in Psychology (Millsap & Maydeu-Olivares, 2009)
Also known as
- Replicated P-technique and dynamic factor analysis
- P-technique
- Dynamic factor analysis
- DFA
- State-space model
- Process factor model
- Intraindividual variation analysis
- Idiographic time-series analysis
- GIMME
- Group Iterative Multiple Model Estimation
- Unified SEM network search
Categories
- Analytics & Behaviour
- Statistics & Modelling