Passing new usevariables when updating MplusObject

[franciscowilhelm]

How can I update the usevariables part of an R object using update function for mplusObject?

I tried it the following way:

m_reg_auto <- map(scalelist, function(y) {

  # create strings
  definevar <-str_c("'", "CENTER", "x", "(GROUPMEAN);", "'", sep = " ")
  modelvar <- str_c("'",y, "ON", "x;","'", sep  = " ")

  # update template
  body <- update(m_reg_template,
  DEFINE = as.formula(paste("~ ",definevar)),
  MODEL = as.formula(paste("~ . + ", modelvar)),
  usevariables = c("id", y, "csm_se"))

  mplusModeler(body, str_c("m_reg_", y, ".dat"), run = TRUE)
 })

However, the usevariables part of the update is not incorporated. Instead, R throws this suggestion:

R variables selected automatically as any variable name that occurs in the MODEL, VARIABLE, or DEFINE section. If any issues, suggest explicitly specifying USEVARIABLES. A starting point may be: USEVARIABLES = id crq_xpl_env csm_se;

[franciscowilhelm]

I can update manually outside the update function through body[["usevariables"]] <- (varnames), but is it possible to update inside the update function?

[JWiley]

@franciscowilhelm

This is a bug caused by the default behavior of trying to guess names running amuck. For now, a work around is:

m1 <- mplusObject(
  MODEL = "mpg ON cyl;",
  rdata = mtcars,
  usevariables = c("mpg", "cyl"),
  autov = FALSE)

m2 <- update(m1,
  MODEL = ~ . + "mpg ON am;",
  usevariables = c("mpg", "cyl", "am", "vs"))

As a quick note, if you are running a variety of models on the same dataset, there is computational efficiency in specifying all the variables you will use to usevariables so that R only has to write the dataset once. That looks like this:

m1 <- mplusObject(
  VARIABLE = "USEVARIABLES = mpg cyl;",
  MODEL = "mpg ON cyl;",
  rdata = mtcars,
  usevariables = c("mpg", "cyl", "am", "vs"),
  autov = FALSE)

m2 <- update(m1,
  VARIABLE = ~ "USEVARIABLES = mpg cyl am;", 
  MODEL = ~ . + "mpg ON am;")

I'm a bit swamped at the moment, but in a week or so I should get a proper fix out (i.e., update will not automatically guess variables regardless of autov flag if usevariables are explicitly passed.

[franciscowilhelm]

Thanks for the heads up! Had not realized that you can use the distinction between Mplus model command USEVARIABLES and usevariables this way. Looking forward to the update.