Announcement

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

  • Margins Fix in January 11 Update for Version 15.1

    I noticed this as part of the update. Does anyone have a sense as to the particular nature of the bug and/or the extent to which other versions are/were affected by it and/or whether specifications using options like dydx(*) were affected?
    b. margins with option expression(), when the expression contained a time-series-operated variable or a
    factor variable and when that same variable specification was also specified in option at(), dydx(),
    dyex(), eydx(), or eyex(), did not correctly compute the requested margin or marginal effect. This has
    been fixed.

  • #2
    Hi John, I ran into a bug using margins with the expression option, when having specified a factor variable in the estimation that I was using in the expression. It was giving me a different message than the one they are mentioning in the update info. This error is not quite what the message in the update that you're quoting is referring to, but it seems to me that expression option has several bugs. I do hope that they get them all.
    Alfonso Sanchez-Penalver

    Comment


    • #3
      Thanks for this input, Alfonso. I use margins, dydx(*) expression(…) with factor variables regularly so am hoping to get clarification on which previous results may be of concern.

      I'm also curious whether using xb specifications in the expression(…)option was also problematic, or whether the factor variables had to be listed explicitly in the expression(…) statement for the bug to arise.

      Comment


      • #4
        I contacted Stata technical support. Here is their reply.

        First of all, thank you for your question about this whatsnew item as it
        has made us realize that this item is actually lacking some precision.
        We will provide a modified version of this in the next update.

        The problem that was fixed was mainly concerning time-series operators.
        Everything said in that item is an accurate description of what has been
        fixed for the cases where you would have a variable with time-series
        operators in the expression and used that variable specification in one
        of the mentioned options. In these cases, the results were incorrect.
        However, the problem was of much smaller scope with respect to factor
        variables. For factor variables, results were only incorrect in the case
        when there was a factor variable specification in the expression (i.e,
        explicitely listed), and the same factor variable specification was used
        in the -at()- option, _and_ if the -at()- values were set to fractional
        values for the levels of that factor variable.

        I hope this clears it up, but please let me know if you have further
        questions about this. I apologize for this whatsnew item not being as
        precise as it should be, and, as I said, we will provide an updated
        version as soon as possible.

        Comment

        Working...
        X