Latent-class mixed models via lcmm

Thin extender over lcmm::lcmm for the Proust-Lima et al. (2017) latent-class linear mixed model on longitudinal / repeated-measures data.

Usage

morie_lcmm_latent_class(fixed, random = ~1, subject, data, ng = 2, ...)

Arguments

fixed

A two-sided formula for the fixed-effects part of the model.

random

A one-sided formula for the random-effects part (default ~1, random intercept only).

subject

Character; the name of the column in data identifying the subject / grouping variable.

data

A data frame containing the variables in fixed, random, and subject.

ng

Integer; the number of latent classes (default 2).

...

Further arguments forwarded to lcmm::lcmm (e.g. mixture, classmb, idiag, nwg, link, intnodes, epsa, epsb, epsd, maxiter, B, convB, convL, convG, verbose).

Value

A list with $method = "lcmm::lcmm" and $raw (an lcmm object with the class-membership probabilities, class-specific fixed-effect estimates, and convergence diagnostics).

Examples

# \donttest{
  if (requireNamespace("lcmm", quietly = TRUE)) {
    data("data_hlme", package = "lcmm")
    # lcmm needs initial values when ng > 1: fit the one-class
    # model first and seed the two-class fit from it (B = ...).
    m1 <- morie_lcmm_latent_class(
      fixed   = Y ~ Time,
      random  = ~ Time,
      subject = "ID",
      data    = data_hlme,
      ng      = 1
    )
    morie_lcmm_latent_class(
      fixed   = Y ~ Time,
      random  = ~ Time,
      subject = "ID",
      data    = data_hlme,
      ng      = 2,
      mixture = ~ Time,
      B       = m1$raw
    )
  }
#> $method
#> [1] "lcmm::lcmm"
#> 
#> $raw
#> General latent class mixed model 
#>      fitted by maximum likelihood method 
#>  
#> lcmm::lcmm(fixed = fixed, mixture = ..1, random = random, subject = subject, 
#>     ng = ng, data = data)
#>  
#> Statistical Model: 
#>      Dataset: data 
#>      Number of subjects: 100 
#>      Number of observations: 326 
#>      Number of latent classes: 2 
#>      Number of parameters: 9  
#>      Link function: linear  
#>  
#> Iteration process: 
#>      Convergence criteria satisfied 
#>      Number of iterations:  18 
#>      Number of iterations:  18 
#>      Convergence criteria: parameters= 1.7e-07 
#>                          : likelihood= 7.2e-08 
#>                          : second derivatives= 6.3e-15 
#>  
#> Goodness-of-fit statistics: 
#>      maximum log-likelihood: -793.83  
#>      AIC: 1605.67  
#>      BIC: 1629.11  
#>  
#>  
#> 
# }