mxsem_group_by {mxsem}

mxsem_group_by

Description

creates a multi-group model from an OpenMx model.

Usage

mxsem_group_by(
  mxModel,
  grouping_variables,
  parameters = c(".*"),
  use_grepl = TRUE
)

Arguments

mxModel

mxModel with the entire data

grouping_variables

Variables used to split the data in groups

parameters

the parameters that should be group specific. By default all parameters are group specific.

use_grepl

if set to TRUE, grepl is used to check which parameters are group specific. For instance, if parameters = "a" and use_grepl = TRUE, all parameters whose label contains the letter "a" will be group specific. If use_grep = FALSE only the parameter that has the label "a" is group specific.

Details

mxsem_group_by creates a multi-group model by splitting the data found in an mxModel object using dplyr's group_by function. The general idea is as follows:

1. The function extracts the data from mxModel 2. The data is split using the group_by function of dplyr with the variables in grouping_variables 3. a separate model is set up for each group. All parameters that match those specified in the parameters argument are group specific

**Warning**: The multi-group model may differ from **lavaan**! For instance, **lavaan** will automatically set the latent variances for all but the first group free if the loadings are fixed to equality. Such automatic procedures are not yet implemented in **mxsem**.

Value

mxModel with multiple groups. Use get_groups to extract the groups

Examples

# THE FOLLOWING EXAMPLE IS ADAPTED FROM
# https://openmx.ssri.psu.edu/docs/OpenMx/latest/_static/Rdoc/mxModel.html
library(mxsem)

model <- 'spatial =~ visual + cubes + paper
          verbal  =~ general + paragrap + sentence
          math    =~ numeric + series + arithmet'

mg_model <- mxsem(model = model,
                  data  = OpenMx::HS.ability.data) |>
  # we want separate models for all combinations of grades and schools:
  mxsem_group_by(grouping_variables = "school") |>
  mxTryHard()

# let's summarize the results:
summarize_multi_group_model(mg_model)