mplusModeler: Create, run, and read Mplus models.

Description

This is a convenience wrapper to automate many of the usual steps required to run an Mplus model. It relies in part on functions from the MplusAutomation package.

Usage

mplusModeler(object, dataout, modelout, run = 0L,
    check = FALSE, varwarnings = TRUE, ...)

Arguments

object

An object of class mplusObject

dataout

the name of the file to output the data to for Mplus.

modelout

the name of the output file for the model. This is the file all the syntax is written to, which becomes the Mplus input file. Typically ends in .inp.

run

an integer indicating how many models should be run. Defaults to zero. If zero, the data and model input files are all created, but the model is not run. This can be useful for seeing how the function works and what setup is done. If one, a basic

check

logical whether the body of the Mplus syntax should be checked for missing semicolons using the parseMplus function. Defaults to FALSE.

varwarnings

A logical whether warnings about variable length should be left, the default, or removed from the output file.

...

additional arguments passed to the prepareMplusData function.

Value

An Mplus model object, with results. If run = 1, returns an invisible list of results from the run of the Mplus model (see readModels from the MplusAutomation package). If run = 0, the function returns a list with two elements, model and boot that are both NULL. if run >= 1,returns a list with two elements, model and boot containing the regular Mplus model output and the boot object, respectively. In all cases, the Mplus data file and input files are created.

Details

Combined with functions from the MplusAutomation package, this function is designed to make it easy to fit Mplus models from R and to ease many of the usual frustrations with Mplus. For example, Mplus has very specific formats it accepts data in, but also very little data management facilities. Using Rdata management is easy. This function is designed to make using data from Rin Mplus models easy. It is also common to want to fit many different models that are slight variants. This can be tedius in Mplus, but using Ryou can create one basic set of input, store it in a vector, and then just modify that (e.g., using regular expressions) and pass it to Mplus. You can even use loops or the *apply constructs to fit the same sort of model with little variants.

Examples

Run this code

# minimal example of a model using builtin data, allowing R
# to automatically guess the correct variables to use
test <- mplusObject(MODEL = "mpg ON wt hp;
  wt WITH hp;", rdata = mtcars)

 # estimate the model in Mplus and read results back into R
 res <- mplusModeler(test, "mtcars.dat", modelout = "model1.inp", run = 1L)

 # show summary
 summary(res)

 # remove files
 unlink("mtcars.dat")
 unlink("model1.inp")
 unlink("model1.out")

# simple example of a model using builtin data
# demonstrates use with a few more sections
test2 <- mplusObject(
  TITLE = "test the MplusAutomation Package and mplusModeler wrapper;",
  MODEL = "
    mpg ON wt hp;
    wt WITH hp;",
  usevariables = c("mpg", "wt", "hp"),
  rdata = mtcars)

 res2 <- mplusModeler(test2, "mtcars.dat", modelout = "model2.inp", run = 1L)

 # remove files
 unlink("mtcars.dat")
 unlink("model2.inp")
 unlink("model2.out")

 # similar example using a robust estimator for standard errors
 # and showing how an existing model can be easily updated and reused
 test3 <- update(test2, ANALYSIS = ~ "ESTIMATOR = MLR;")

 res3 <- mplusModeler(test3, "mtcars.dat", modelout = "model3.inp", run = 1L)
 unlink("mtcars.dat")
 unlink("model3.inp")
 unlink("model3.out")

 # now use the built in bootstrapping methods
 # note that these work, even when Mplus will not bootstrap
 # also note how categorical variables and weights are declared
 # in particular, the usevariables for Mplus must be specified
 # because mroe variables are included in the data than are in the
 # model. Note the R usevariables includes all variables for both
 # model and weights. The same is true for clustering.
 test4 <- mplusObject(
   TITLE = "test bootstrapping;",
   VARIABLE = "
     CATEGORICAL = cyl;
     WEIGHT = wt;
     USEVARIABLES = cyl mpg;",
   ANALYSIS = "ESTIMATOR = MLR;",
   MODEL = "
     cyl ON mpg;",
   usevariables = c("mpg", "wt", "cyl"),
   rdata = mtcars)

 res4 <- mplusModeler(test4, "mtcars.dat", modelout = "model4.inp", run = 10L)
 # see the results
 res4$results$boot

 # remove files
 unlink("mtcars.dat")
 unlink("model4.inp")
 unlink("model4.out")
 unlink("Mplus Run Models.log")