Announcement

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

  • lclogit/lclogitml: change of reference latent class

    Dear all,

    I am using the lclogit/lclogitml command in Stata/SE 13.1 with 3 or more latent classes. I am especially interested in the t-statistics of the class membership model

    - Is there a way to change the reference class for the class membership model in lclogit/lclogitml?
    or
    - If changing reference class is not possible, is there a way to calculate standard errors for a manuel change of the reference class. I cannot see that the variance/covariance matrix of the membership model is part of the lclogitml output.

    Thank you very much
    Gerald

  • #2
    Gerald

    Remember that the Statalist FAQ asks you to say if a program is user-written. lclogit (SJ, SSC) is written by Daniele Pacifico and Hong Il Yoo.

    There is no need to change the reference category - you can calculate all of the class membership probabilities (and their standard errors) using nlcom, e.g.:

    Code:
    use http://fmwww.bc.edu/repec/bocode/t/traindata.dta, clear
    lclogit y price contract local wknown tod seasonal, id(pid) gr(gid) ncl(3)
    lclogitml, iterate(10)
    nlcom (p_class1: exp([share1]_cons)/(exp([share1]_cons)+exp([share2]_cons)+1))
    nlcom (p_class2: exp([share2]_cons)/(exp([share1]_cons)+exp([share2]_cons)+1))
    nlcom (p_class3: 1/(exp([share1]_cons)+exp([share2]_cons)+1))
    Arne

    Comment


    • #3
      Dear Prof. Hole,

      My point is similar to the one raised by Gerald.
      I am using the lclogit to investigate job preferences among doctors. lclogit is a user-written stata program written by Daniele Pacifico and Hong Il Yoo.
      I use the results of a DCE with some usual attributes. I try to explain segmentation of doctors into classes using a residency exam_score (e.g. USMLE)

      Code:
      Code:
       
      lclogit Choice Income Hours Place Night_ED PHC Work Private Research, group(GID) id(ID) nclasses(4) membership(exam_score)
      If I am understanding the model correctly, the coefficients for the exam_score in each class give me who were the best/worst doctors according to this exam and segment job preferences accordingly (say job preferences for "high achievers", job preferences for "low achievers").
      I would like to set the reference class to the lowest score class. I think this would be easier to interpret : "how groups with increasing scores in the exam respond to changes in income, or changes in hours, etc."

      Thanks
      Pedro

      Comment

      Working...
      X