latents-model-methods: Coefficients, covariance, intervals and likelihood of latents...

In latents: Multilevel Latent Profile, Class and Transition Models

Coefficients, covariance, intervals and likelihood of latents fits

Description

coef(), vcov(), confint(), logLik() and nobs() for multilevel profile and class models (multilpa()), covariate models, latent transition models and, for vcov(), pooled imputations. Mixture regressions, growth mixtures and the additive, cross-level and general transition families document the same methods with their get_results() page.

Usage

## S3 method for class 'multilpa_covariates'
vcov(
  object,
  data = NULL,
  step = 1e-04,
  vcov_type = c("observed", "robust", "opg"),
  scale = c("natural", "unconstrained"),
  boundary = c("error", "fix"),
  ...
)

## S3 method for class 'multilpa_covariates'
coef(object, scale = c("natural", "unconstrained"), ...)

## S3 method for class 'multilpa_covariates'
confint(object, parm, level = 0.95, data = NULL, ...)

## S3 method for class 'multilpa_covariates'
logLik(object, ...)

## S3 method for class 'multilpa_covariates'
nobs(object, ...)

## S3 method for class 'multilpa'
coef(object, scale = c("natural", "unconstrained"), ...)

## S3 method for class 'multilpa'
vcov(object, data = NULL, scale = c("natural", "unconstrained"), ...)

## S3 method for class 'multilpa'
confint(object, parm, level = 0.95, data = NULL, ...)

## S3 method for class 'multilpa'
logLik(object, ...)

## S3 method for class 'multilpa'
nobs(object, ...)

## S3 method for class 'latents_pooled'
vcov(object, ...)

## S3 method for class 'multilpa_transitions'
logLik(object, ...)

## S3 method for class 'multilpa_transitions'
nobs(object, ...)

## S3 method for class 'multilpa_transitions'
coef(object, ...)

Arguments

object	A fitted model: from multilpa() (class multilpa), a covariate model (multilpa_covariates), a latent transition model (multilpa_transitions), or pooled imputations (latents_pooled, vcov() only).
data	Optional. The data frame the model was fitted to. When omitted it is rebuilt from what the fit stores (indicators, identifiers, occasions and designs), which round-trips exactly; supplying it is the stronger check that the caller still holds that frame.
step	Finite-difference step for the observed information (covariate models).
vcov_type	"observed", "robust" or "opg", as for parameter_inference() (covariate models).
scale	"natural", the default, or "unconstrained". Natural coefficients are in the units coef() reports: means and variances in the indicators' units, probabilities as probabilities, with the covariance carried from the estimation scale by the delta method. Unconstrained is the scale the model is estimated on: log variances (diagonal), log-Cholesky coordinates (full covariance), log volumes, shapes and orientations (a structure that constrains them across profiles), and baseline-category logits for mixing and response probabilities. In a covariate model, means and membership coefficients are the same on both scales.
boundary	"error" or "fix", as for parameter_inference(): "fix" holds response probabilities on their bound, whose rows and columns are then zero (covariate models).
...	For vcov() and confint() of a multilpa fit, and confint() of a covariate model, passed to parameter_inference() (for example method = "bootstrap", vcov_type = "robust" or step). Ignored otherwise.
parm	Optional coefficient names or indices; defaults to every estimated coefficient. Naming a coefficient that fixed held raises latents_held_parameter, because a held value has no interval.
level	Confidence level strictly between zero and one.

Value

coef(): a named numeric vector. vcov(): a named symmetric matrix over the same parameters. confint(): a matrix with one row per parameter and the lower and upper bounds as columns. logLik(): a "logLik" object with df and nobs attributes. nobs(): the number of groups, the independent units of the likelihood.

See Also

parameter_inference() for the tidy table of estimates, standard errors and intervals.

Examples

fit <- multilpa(course_engagement, c("browse", "lectures", "forum_read"),
                "student", n_profiles = 2, n_group_classes = 2,
                n_starts = 2, seed = 1)
coef(fit)
logLik(fit)
nobs(fit)

latents documentation built on Oct. 16, 2026, 5:09 p.m.