DDDFI: Direct Discrepancy Dynamic fit index (3DFI) cutoffs for...

In melissagwolf/dynamic: DFI Cutoffs for Latent Variable Models

Direct Discrepancy Dynamic fit index (3DFI) cutoffs for arbitrary covariance structure models

Description

This function generates DFI cutoffs for any single group covariance structure model with a saturated or absent mean structure. It supports (a) any estimator supported by lavaan (e.g. ML, MLR, WLSMV, ULSMV), (b) missing data, and (c) multiple response scales (normal, non-normal continuous, categorical). The default argument is a singular argument: a lavaan object.The function can also accommodate manual entry of the model statement and sample size. Some features require an original dataset to be provided (e.g., missing data, categorical data). The app-based version of this function can be found at dynamicfit.app.

Usage

DDDFI(
  model,
  data = NULL,
  scale = "normal",
  manual = FALSE,
  reps = 250,
  n = NULL,
  estimator = NULL,
  MAD = c(0.038, 0.05, 0.06),
  plot.dfi = FALSE,
  plot.dist = FALSE,
  plot.discrepancy = FALSE
)

## S3 method for class 'DDDFI'
print(x, ...)

Arguments

model	This can either be a lavaan object, OR a model statement written in lavaan model.syntax with standardized estimates
data	The original data to which the model was applied. Not required if scale="normal". Otherwise, data is required.
scale	Determines how data are simulated. Options are "normal", "nonnormal", or "categorical". "normal" assumes multivariate normality across all variables. "nonnormal" recreates distributions in an empirical dataset (to be provided by the user), assuming variables are continuous. "categorical" simulates discrete data with the same proportions as an empirical dataset (to be provided by the user). With "categorical", mixed formats are also supported and any variable with more than 9 categories is simulated from a normal distribution.Only "normal" can be used without provided an original dataset.
manual	If you entered a lavaan object, keep this set to FALSE. If you manually entered standardized estimates and sample size, set this to TRUE.
reps	The number of replications used in the simulations. This is set to 250 by default
n	If you entered a lavaan object for model, leave this blank. Otherwise, enter your sample size (numeric).
estimator	Which estimator to use within the simulations (enter in quotes). The default depends on the scale option ("ML" for "normal", "MLR" for "nonnormal", and "WLSMV" for categorical)
MAD	Mean Absolute Discrepancies to test in the simulation. Default is c(.038, .05, .06) to recreate traditional "Close", "Fair", "Mediocre" benchmarks
plot.dfi	Displays simulated distributions of fit indices used to derive cutoffs for each MAD value.
plot.dist	Displays distributions of simulated data (and empirical data, if provided) to assess fidelity of simulated data to empirical data
plot.discrepancy	Displays distributions of simulated MAD values
x	DDDFI object
...	other print parameters

Value

Direct Discrepancy Dynamic fit index (DFI) cutoffs for CFI, RMSEA, and RMSEA 90

Author(s)

Daniel McNeish & Melissa G Wolf

Maintainer: Daniel McNeish <[e-mail]>

Examples

#Example using a lavaan object as input (manual=FALSE)

lavmod <- "F1 =~ x1 + x2 + x3
           F2 =~ x4 + x5 + x6
           F3 =~ x7 + x8 + x9"

fit <- lavaan::cfa(lavmod,data=Holzinger)
DDDFI(fit)

#Manual entry example (manual=TRUE)

#Holzinger 3-factor model
lavmod <- "F1 =~ x1 + x2 + x3
           F2 =~ x4 + x5 + x6
           F3 =~ x7 + x8 + x9"

 #fit the model,lavaan is used here to shown where estimates come from
 #but manual entry supports standardized estimates from models fit in any software

 fit<-lavaan::cfa(lavmod, data=Holzinger)
 lavaan::standardizedsolution(fit)

 #model statement with standardized estimates
 manual_model <- "F1 =~ .772*x1 + .424*x2 + .581*x3
            F2 =~ .852*x4 + .855*x5 + .838*x6
            F3 =~ .570*x7 + .723*x8 + .665*x9
            F1~~.459*F2
            F1~~.471*F3
            F2~~.283*F3"

DDDFI(model=manual_model,n=301,manual=TRUE)

melissagwolf/dynamic documentation built on June 29, 2024, 6:24 p.m.