Create Thresholds Based on Deviances — mx

This function creates a list of mxMatrices to specify thresholds for ordinal variables as a function of strictly positive deviances, see Details.

Usage

mx_deviances(
  vars,
  nThresh = NA,
  free = NULL,
  values = NULL,
  labels = NA,
  lbound = NULL,
  ubound = NULL
)

Arguments

vars: Character vector. Names of variables for which thresholds should be specified.
nThresh: Numeric vector. Number of thresholds for each of variable in vars.
free: Logical vector, used to fill the free matrix which indicates whether the deviance will be freely estimated or fixed.
values: Numeric vector of starting values. Default uses mxNormalQuantiles().
labels: Character vector. Note that constraints specified using these labels are applied using mxConstraint(), because the thresholds are derived quantities, not parameters of the model (the deviances are).
lbound: PARAM_DESCRIPTION, Default: NULL
ubound: PARAM_DESCRIPTION, Default: NULL

Value

A list of mxMatrix objects, to be included in an mxModel.

Details

Specifying thresholds as a function of strictly positive deviances is an efficient way to get models with ordinal indicators to converge, if they tend to arrive at impossible solutions where the thresholds are not monotonously increasing (e.g., in mixture models). The parameters of a model specified this way are deviances (i.e., the differences between threshold values), which have a lower bound of a small positive number to ensure that thresholds are always increasing. The thresholds are computed by adding subsequent deviances.