lessSEM with an unfitted lavaan model

[B260198]

Dear Jannik,

I wanted to ask you if it is possible to specify an unfitted lavaan model in the argument "lavaanModel = " in the lasso function of lessSEM. If I do this like here in the example:

lassoFit <- lasso (lavaanModel = unfittedlavaan, data = data, 
                  regularized = c("deltaa5", "deltaa6", "deltaa7"), 
                  nLambdas = 100,
                  method = "glmnet",
                  control = controlGlmnet(),
                  modifyModel = modifyModel(transformations = transformations)
)

I always get the error message " unused argument (data = data)". Does the lasso function not know the "data = " argument or do I perhaps have to specify the data differently? Or is it simply not possible to enter an unfitted model? I thought maybe it works because lasso refits the model again anyway...

Thank you and best regards, Britta

[jhorzek]

Hi Britta, great question! lessSEM expects that the model was fitted with lavaan. However, what you can do is to not let lavaan optimize the model:

library(lessSEM)

dataset <- simulateExampleData()

lavaanSyntax <- "
f =~ l1*y1 + l2*y2 + l3*y3 + l4*y4 + l5*y5 + 
     l6*y6 + l7*y7 + l8*y8 + l9*y9 + l10*y10 + 
     l11*y11 + l12*y12 + l13*y13 + l14*y14 + l15*y15
f ~~ 1*f
"

lavaanModel <- lavaan::sem(lavaanSyntax,
                           data = dataset,
                           meanstructure = TRUE,
                           std.lv = TRUE,
                           do.fit = FALSE # don't optimize, just set up the model; see ?lavOptions 
                           # for more details
                           )

lsem <- lasso(
  lavaanModel = lavaanModel,
  regularized = paste0("l", 6:15),
  nLambdas = 50)

All arguments of the lasso - function are explained here: https://jhorzek.github.io/lessSEM/reference/lasso.html#arguments.

Best, Jannik