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

  • When convergence is not achieved with time dummies

    I will appreciate guidance with this.
    I have a panel data from 41 states over 40 time periods. My dependent variable is binary (adopt or not adopt). I used Hausman test and determined that fixed effect is preferred. I am seeking to use the conditional fixed effects logistic regression to estimate my odds ratios. I generated 40 time periods and have included 39 time dummies. After running the model for 24 HOURS no convergence is achieved and had to stop it. I find that the regression is successful without time dummies, or when I exclude the last 20 years, or select alternative years but not the last few years. I find few of the time variables significant.
    Two peer reviewers asked I include time dummies. Any advice?


  • #2
    welcome to this forum.
    Instead of creating time dummies yourself, why not relying on the wonderful capabilities of -fvvarlist- notation?
    I cannot say if this will speed your calculation up, but surely you will benefit from a straight relationships with -margins- and -marginsplot-, extermely useful in post estimation.
    Last edited by Carlo Lazzaro; 18 May 2018, 01:05.
    Kind regards,
    (Stata 15.1 SE)