Announcement

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

  • Heckman not concave, but probit converge does it

    Hi
    I have a strange problem.
    When I try to run a model using heckman my model is not concave, however when I make the heckman model in manual mode (probit and then reg including the IMR, following https://www.stata.com/statalist/arch.../msg00728.html ) I have no problems.
    In addition, in the probit model, iterations concludes when it reach the same log likelihood in which the Heckman model announces that it is not concave.
    It's not clear to me why the discrepancy

    This is my log (in a picture format)




  • #2
    Hi Ricardo,
    the picture you send didnt attached correctly.
    A more helpful advice could be given if you provide us with the exact code that you are using. My best guess is that you are comparing heckman ML with heckman twostep.
    By default, heckman estimates the ml model, and if you want it to estimate something like what you did manually, you need to use the option "two"
    HTH
    Fernando

    Comment


    • #3
      Hii
      Thank you for the reply.

      My sintax is

      "probit ss_dum $agro1 $agro2 $demo1 $demo2 $other1 $other2 i.id_coop i.dist if year == 0 & dist_coop != . & tiempovp_pie !=.
      predict phat, xb
      gen mills=exp(-0.5*phat*phat)/(sqrt(2*_pi)*normprob(phat))
      reg pp_ss_vta_tot_heck $agro1 $agro2 $demo1 $demo2 $other1 $other2 i.id_coop i.dist mills if year == 0 & dist_coop != . & tiempovp_pie !=.

      heckman pp_ss_vta_tot_heck2 $agro1 $agro2 $demo1 $demo2 $other1 $other2 i.id_coop i.dist if year == 0 & dist_coop != . & tiempovp_pie !=. , ///
      select(ss_dum= $agro1 $agro2 $demo1 $demo2 $other1 $other2 i.id_coop i.dist ) first twostep
      "

      Comment


      • #4
        can you also provide summary of pp_ss_vta_tot_heck2 and pp_ss_vta_tot_heck for ss_dum==0 and ss_dum==1.
        and the screen shot of the problem you are facing? I find it difficult to see what is happening that it does not converge with the heckman two step syntax. The summary Im asking may help to understand what may be happening.
        Fernando.

        Comment


        • #5
          Thank you for the reply.
          Since i have problems attaching the images, i have uploaded the pictures in a external server.
          The first one wiht the summary that you ask and the second with the problem


          https://imgur.com/8YbDFMl
          https://imgur.com/4ULi1b3


          Thanks for your time

          Comment

          Working...
          X