med.fit.BRR: Estimate the mediation effects and standard errors adjusting...

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

Description

This function is used to estimate the mediation effects adjusted for complex surveys with balanced repeated replications (BRR) (Mai, Ha, Soulakova, 2019).

Usage

med.fit.BRR(
  model = NULL,
  data = NULL,
  mwgtname = NULL,
  repwgtnames = NULL,
  fayfactor = 0.5,
  estimator = c("ML"),
  test = c("satorra.bentler"),
  parallel = c("no", "parallel", "snow"),
  ncore = Sys.getenv("NUMBER_OF_PROCESSORS"),
  cl = NULL,
  ...
)

Arguments

model	The model being fitted. It is written in lavaan model syntax (Rosseel, 2012).
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.
parallel	Parallel computing ("no" or "parallel" or "snow"). It is "no" by default, which means it will not use parallel computing. The option "parallel" is to use multiple cores in a computer for parallel computing. It is used with the number of cores (ncore). The option "snow" is to use clusters for parallel computing. It is used with the number of clusters (cl).
ncore	Number of processors used for parallel computing. By default, ncore = Sys.getenv ('NUMBER_OF_PROCESSORS').
cl	Number of clusters. It is NULL by default. When it is NULL, the program will detect the number of clusters automatically.
...	Extra arguments to be incorporated with in the future. It is not required.

Value

The model fit results as a lavaan object with the adjusted estimates, standard errors, and model fit statistics. It is a lavaan object (Rosseel, 2012).

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.

Mai, Y., Ha, T., & Soulakova, J. N. (2019). Multimediation Method With Balanced Repeated Replications For Analysis Of Complex Surveys. Structural Equation Modeling: A Multidisciplinary Journal. DOI:10.1080/10705511.2018.1559065

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

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)]

model2 <- ' # 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
            #covariance of residuals
              sp_adltban ~~ sp_kidsban
            # indirect effect (a*b)
              a1b1 := a1*b1
              a2b2 := a2*b2
            # total effect
              total := c + (a1*b1) + (a2*b2)
           '
fit.BRR <- med.fit.BRR(model=model2, data=MedData, mwgtname=mwgtname,
         repwgtnames=repwgtnames, fayfactor=0.5, parallel='parallel', ncore=2)
lavaan::summary(fit.BRR)
#
# lavaan 0.6-3 ended normally after 41 iterations
#
# Optimization method                           NLMINB
# Number of free parameters                         12
#
# Number of observations                          3922
#
# Estimator                                         ML      Robust
# Model Fit Test Statistic                       0.000       0.000
# Degrees of freedom                                 0           0
# Minimum Function Value               0.0000000000000
# Scaling correction factor                                     NA
# for the Satorra-Bentler correction
#
# Parameter Estimates:
#
#   Information                                 Expected
# Information saturated (h1) model          Structured
# Standard Errors                                  BRR
#
# Regressions:
#                    Estimate  Std.Err  z-value  P(>|z|)
# numcg ~
#    workban    (c)   -0.101    0.039   -2.572    0.010
#    sp_adltbn (b1)   -0.253    0.048   -5.270    0.000
#    sp_kidsbn (b2)   -0.361    0.051   -7.006    0.000
# sp_adltban ~
#    workban   (a1)    0.069    0.018    3.753    0.000
# sp_kidsban ~
#    workban   (a2)    0.020    0.016    1.250    0.211
#
# Covariances:
#                    Estimate  Std.Err  z-value  P(>|z|)
# .sp_adltban ~~
#    .sp_kidsban        2.784    0.195   14.300    0.000
#
# Intercepts:
#                    Estimate  Std.Err  z-value  P(>|z|)
#   .numcg     (u0)   18.485    0.566   32.668    0.000
#   .sp_adltbn (u1)    4.221    0.167   25.281    0.000
#   .sp_kidsbn (u2)    7.926    0.143   55.272    0.000
#
# Variances:
#                    Estimate  Std.Err  z-value  P(>|z|)
#   .numcg            54.283    1.716   31.628    0.000
#   .sp_adltban       11.011    0.239   46.140    0.000
#   .sp_kidsban        9.402    0.209   44.998    0.000
#
# Defined Parameters:
#                    Estimate  Std.Err  z-value  P(>|z|)
#    a1b1             -0.017    0.006   -2.905    0.004
#    a2b2             -0.007    0.006   -1.234    0.217
#    total            -0.125    0.040   -3.169    0.002

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