Fits a stepwise measurement-invariance model and then an HMM whose latent states represent clusters of groups over time. Each group must be observed at every time point, with at least two time points.
Usage
lmmgsem(
dat,
S1 = NULL,
S2,
group,
time,
nclus,
seed = NULL,
max_it = 10000L,
nstarts = 20L,
printing = FALSE,
s2_fast = TRUE,
s1_fit = NULL,
mi_method = c("group_violations", "test_statistic"),
mi_alpha = 0.05,
mi_d_cfi_threshold = 0.01,
mi_max_iter = 20L,
mi_fit_indices = c("chisq", "df", "pvalue", "cfi", "rmsea", "srmr"),
mi_specific_noninv = FALSE,
...
)Arguments
- dat
Data frame containing the observed variables, group identifier, and time identifier.
- S1
Measurement model in lavaan syntax. Required when
s1_fitis not supplied.- S2
Structural model in lavaan syntax.
- group
Name of the grouping variable in
dat.- time
Name of the time variable in
dat.- nclus
Number of latent states (clusters).
- seed
Optional random seed for reproducibility.
- max_it
Maximum number of EM iterations per random start.
- nstarts
Number of random starts.
- printing
If
TRUE, print iteration progress.- s2_fast
If
TRUE, use the fast Step 2 estimator; otherwise use the slow estimator.- s1_fit
Optional lavaan measurement-model fit with one group per group-time cell. Its group labels must match
interaction(dat[[group]], dat[[time]], sep = ".Time"). If omitted, measurement invariance is estimated usingcompute_mi_long_stepwise_score().- mi_method
Method for selecting invariance violations:
"group_violations"or"test_statistic".- mi_alpha
Significance level for the score tests.
- mi_d_cfi_threshold
Maximum configural-to-metric CFI difference.
- mi_max_iter
Maximum iterations in each stepwise invariance analysis.
- mi_fit_indices
Fit indices returned by the invariance analysis.
- mi_specific_noninv
If
TRUE, allow non-invariance for specific group-time pairs; otherwise free loadings globally.- ...
Additional arguments passed to lavaan's CFA fitting functions during the measurement-invariance analysis.
Value
A lmmgsem object containing time-specific posterior state
probabilities, transition probabilities, model parameters, likelihoods,
and model-selection criteria. Information criteria use the sequence-level
factor likelihood and marginal posterior state entropy. Standard errors
are not currently implemented. Parameter counts include free measurement
parameters, group-time-specific exogenous and endogenous covariance
parameters, per-state regression parameters, and initial/transition
probabilities.