CMSE: CMSE: Compute the Causal Mean Squared Error

In trbrick/CMSE: CMSE: Causal Mean Squared Error index

Description

'Causal Mean Squared Error (CMSE)' is a fit index that quantifies misfit between the model-implied causal effects of an intervention and the empirically determined effects.

CMSE computes the Causal Mean Squared Error (CMSE). CMSE indexes the difference between the expected effect of the intervention and the real impact of the intervention. Specifically, the CMSE computes the true effect of the intervention on each outcome using a simple linear regression (controlling appropriately for covariates) and compares that true effect to the model-implied effect on that outcome. By focusing on the specific causal predictions of the model, CMSE captures a different slice of information than other fit statistics.

See Wan, et al., for more.

Usage

CMSE(
  dataset,
  intervention,
  models,
  posttest,
  outcomes,
  covariates = NULL,
  ...,
  nrows = NA,
  latentPosttest = NA
)

Arguments

dataset	the actual data as a data frame or mxDataStatic object
intervention	character string containing the name of the intervention column
models	a list of mxModel objects (like those returned from makeARPanelModels) for fitting
posttest	character string containing the name of the posttest outcome
outcomes	character vector including names of all additional outcomes
covariates	character vector including the names of all covariates to be included
...	Does not accept arguments; only there so that later arguments must be named
nrows	number of data rows in the data set; Required if dataset is a covariance matrix without means
latentPosttest	the latent true posttest score; Required if the posttest outcome has a measurement model (e.g. for the RI-AR model)

Details

This is quantified as the difference between the (regression-determined) empirical effects of the intervention on distal outcomes, and the model's predictions about those distal effects given the empirical proximal effect. See details for more.

CMSE: Causal Mean Squared Error index

Value

a list containing elements: - '$CMSE' a named data frame of CMSE scores for each model and each outcome, as well as model-mean CMSEs - '$experimental' contains the experimentally-determined effect of the intervention on posttest and outcomes - '$nonexperimental' contains the model-implied effect of the intervention on posttest and each outcome

trbrick/CMSE documentation built on March 17, 2021, 4:15 a.m.