No announcement yet.
  • Filter
  • Time
  • Show
Clear All
new posts

  • Latent class models using lclogit & lclogitml

    I am trying to estimate a two-class latent class modeling using lclogit from ssc and its post-estimation command lclogitml (which uses gllamm from ssc) on Stata 14.2.

    The model converges when no weights are used, but when I try to use the surveyweights, e.g.

    lclogit choice $attributes, group(choiceid)
    lclogitml,iterate(80) pweight(pweight)
    where my survey weight variable name is "pweight2", I receive the following error messages:
    TwoStep: did not reach goal
    TwoStep: did not reach goal
    could not calculate numerical derivatives
    missing values encountered
    something went wrong in comprob3
    Has anyone else encountered problems with using survey weights with lclogitml or gllamm, or have suggestions?