clusterwiselmm: Geometrical clusterwise method for linear multi level models

In Redha-ALLICHE/clusterwiselmm: Apply geometrical clusterwise to linear multi-level models (lmm)

Description

This function fits a clusterwise multi level model with k clusters to a given data frame using a geometrical approach. It runs the optimizer for many trials to get rid of the initialization bias and reach a global minima. You can choose between a batch or sequential approach and also fix the initial clustering.

Usage

clusterwiselmm(
  data,
  target,
  K,
  formula,
  nb_trials = 50,
  nb_iterations = 100,
  batch = TRUE,
  verbose = TRUE,
  initialization = NULL
)

Arguments

data	data frame containing the variables named in formula.
target	vector containing the target variable.
K	number of clusters.
formula	formula a two sided linear formula object corresponding to formula parameter in lmer from lme4 package.
nb_trials	number of trials.
nb_iterations	maximum number of iterations.
batch	factor; if TRUE, runs a batch approach, otherwise runs a sequential one.
verbose	factor; if TRUE, prints the final cost at each trial.
initialization	if vector of the same length of data, use this vector as clustering initialization. Otherwise, do a random initialization.

Value

list containing models as a class of clmm, the final clustering and the final cost.

Examples

require(stats)
## define correlation matrices for random effects slopes and intercepts
lvl2_corr_mat = matrix(c(1, 0,
                         0, 1), 2, 2)
lvl3_corr_mat = matrix(c(10, 0,
                         0, 10), 2, 2)

## generate 3 levels data with 3 groups that contains each 10 subgroups,
## and each subgroup contains 30 individuals
data = simulate_data(n1 = 30,
                     n2 = 10,
                     n3 = 3,
                     fixed_slope = -2,
                     fixed_intercept = 2,
                     lvl2_corr_mat = lvl2_corr_mat,
                     lvl3_corr_mat = lvl3_corr_mat,
                     sigma2 = 1
                     )
## fit a clusterwise lmm with 3 classes that tries to find level 3 clustering.

Redha-ALLICHE/clusterwiselmm documentation built on Aug. 3, 2020, 1:10 a.m.