chisq.BRR: Adjust the model fit statistics

In MedSurvey: Linear Mediation Analysis for Complex Surveys Using Balanced Repeated Replication

Description

This function is used to adjust model fit statistics for complex surveys with balanced repeated replications (Oberski, 2014; Satorra & Muthen, 1995). It saves time to only obtain the model fit statistics during the model selection stage.

Usage

chisq.BRR(
  model,
  lavaan.fit,
  data,
  mwgtname,
  repwgtnames,
  fayfactor = 0.5,
  estimator = c("ML"),
  test = c("satorra.bentler")
)

Arguments

model	The model being fitted. It is written in lavaan model syntax (Rosseel, 2012).
lavaan.fit	The model fit results using 'ML' estimator with sample main weights, but without adjusting the fit statistics or standard errors for complex surveys. Note that it is a lavaan object.
data	The raw data including the variables of interest and the survey weights. It should be a dataset or dataframe.
mwgtname	The variable name indicating the sample main weight in the dataset. See balanced repeated replications method (Wolter, 2007) for more information about the main weight.
repwgtnames	The variable names indicating the set of replicate weights in the dataset. See balanced repeated replications method (Wolter, 2007) for more information about the replicate weight.
fayfactor	The fayfactor used in the standard error calculation by fay's method (Fay & Train, 1995; Judkins, 1990) for balanced repeated replications. Fayfactor is a value between 0 and 1. The default is 0.5.
estimator	The method used to estimate the model. 'ML' is the default option and the only available option for current version. It is not required.
test	The method used to generate adjusted standard errors. 'satorra.bentler' is the default option and the only available option for current version. It is not required.

Value

The model fit results as a lavaan object (Rosseel, 2012) with the adjusted model fit statistics.

References

Fay, R. E., & Train, G. F. (1995). Aspects of survey and model-based postcensal estimation of income and poverty characteristics for states and counties. In Proceedings of the Section on Government Statistics, American Statistical Association, Alexandria, VA (pp. 154-159).

Judkins, D. R. (1990). Fay’s method for variance estimation.Journal of Official Statistics,6(3), 223-239.

Oberski, D. (2014). lavaan. survey: An R package for complex survey analysis of structural equation models. Journal of Statistical Software, 57(1), 1-27. DOI:10.18637/jss.v057.i01

Rosseel, Y. (2012). Lavaan: An R package for structural equation modeling and more. Version 0.5–12 (BETA). Journal of statistical software, 48(2), 1-36. DOI:10.18637/jss.v048.i02

Satorra, A., & Muthen, B. (1995). Complex sample data in structural equation modeling. Sociological methodology, 25(1), 267-316.

Wolter, K. (2007). Introduction to variance estimation. New York, NY: Springer.

Examples

R <- 160
wgtnames <- paste("repwgt", seq(0,R,by=1), sep="")
mwgtname=wgtnames[1]
repwgtnames=wgtnames[2:(R+1)]
fayfactor=0.5

model3 <- ' # outcome
            numcg ~ u0*1 + c*workban + b1*sp_adltban + b2*sp_kidsban
           # mediator
              sp_adltban ~ u1*1 + a1*workban
              sp_kidsban ~ u2*1 + a2*workban
           # indirect effect (a*b)
              a1b1 := a1*b1
             a2b2 := a2*b2
           # total effect
             total := c + (a1*b1) + (a2*b2)
          '

fit <- lavaan::sem(model=model3, data=MedData, estimator='ML', test='standard')
chisq.BRR(model3,fit,MedData,mwgtname, repwgtnames)
#
# MedSurvey 1.1.0 Adjusted Model Fit Statistics using BRR
#
# chisq   df   pvalue    CFI      RMSEA      SRMR         AIC       BIC
#
# 305.25   1  0.00000   0.40561  0.27852   0.07416   88699.43   88768.45

MedSurvey documentation built on Oct. 23, 2020, 7:15 p.m.