daob/lavaan.survey

Complex sampling analysis of structural equation models

lavaan.survey's Issues

Deprecation of lavaan::duplicationMatrix

Hi.
I am having difficulty using the lavaan.survey function. It seems that it is using a deprecated lavaan function, so I'm getting an error when I run it.

Warning message:
'lavaan::duplicationMatrix' is deprecated.
Use 'lav_matrix_duplication' instead.
See help("Deprecated") and help("lavaan-deprecated").

More information is in the Lavaan Google Group: https://groups.google.com/forum/#!topic/lavaan/yEUcJEWAPq4

model not found: using lavaan.survey() inside a function environment

Because I need to fit multiple SEMs with weights, I decided to make a convenience function for this. However, it seems that lavaan.survey() fails to work inside a function environment. My function:

sem_weights = function(model_str, data, weights) {
  #fit without weights
  fit = sem(model_str, data = data)

  #with weights
  design = svydesign(ids = ~1,
                     data = data,
                     weights = weights)

  fit_wtd = lavaan.survey(fit, design)
  fit_wtd
}

Calling this with the appropriate parameters:

mod = "Sepal.Length ~ Sepal.Width"
fit = sem_weights(model_str = mod, data = iris, weights = runif(n = nrow(iris), min = 1, max = 2))

Gives a not found error:

Error in lavaan::lavaan(model = model_str, model.type = "sem", int.ov.free = TRUE,  : 
  object 'model_str' not found

I have tried renaming the object but that didn't help. My guess is that lavaan is searching for the model object in the wrong environment.

Running the same code not using the function works fine:

#fit without weights
fit = sem(model = mod, data = iris)

#with weights
design = svydesign(ids = ~1,
                   data = iris,
                   weights = runif(n = nrow(iris), min = 1, max = 2))

fit_wtd = lavaan.survey(fit, design)
#no errors

One can also make the function work by making sure that the model object exists under that name in the global environment, but that defeats the purpose of using a function.

Ideas?