Announcement

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

  • Regressions for subsets

    Hi all,

    I have been trying to conduct regression analysis for two different groups. Later on, I want to compare the coefficients and see whether they are statistically different.
    Now I am using Stata 13.1 and the following code:

    Code:
    xi: areg varlist i.year, absorb(industry) robust cluster(company_id)
    The xi: in combination with i.year seem to be troublesome as it is not possible to use the 'if' qualifier.

    What is the best way to run these regressions for different groups and to be able to compare them afterwards?

    Thanks in advance,

    Jules

  • #2
    Jules:
    welcome to the list.
    -xi- is pleonastic if your Stata version allows you to exploit the wonderful capabilities of -fvvarlist- (as it should be the case, since you are assumed to use Stata 14. Otherwise, you should state which Stata release supports your statistical analysis).
    If -industry- categories are not too many, you may want to switch to -regress- with -if- qualifier and then compare the coefficients across the two models via -suest- (that, unfortunately, does not support -areg-).
    Kind regards,
    Carlo
    (Stata 19.0)

    Comment


    • #3
      Another approach is to include an terms interacting your key predictors with the grouping variable in a single regression model and then look at the interaction results.

      Comment


      • #4
        Originally posted by Carlo Lazzaro View Post
        Jules:
        welcome to the list.
        -xi- is pleonastic if your Stata version allows you to exploit the wonderful capabilities of -fvvarlist- (as it should be the case, since you are assumed to use Stata 14. Otherwise, you should state which Stata release supports your statistical analysis).
        If -industry- categories are not too many, you may want to switch to -regress- with -if- qualifier and then compare the coefficients across the two models via -suest- (that, unfortunately, does not support -areg-).
        Thanks for the swift and helpful response. As stated, I use Stata 13, but it seems to support -fvvarlist-; very helpful, thanks. But I have more than 180 industries, thus running individual regressions is going to be rather tiresome and messy.

        Originally posted by Clyde Schechter View Post
        Another approach is to include an terms interacting your key predictors with the grouping variable in a single regression model and then look at the interaction results.
        Thanks, of course, good point. I am going to try this one.

        Comment


        • #5
          Jules:
          go as per Clyde's helpful suggestion.
          Kind regards,
          Carlo
          (Stata 19.0)

          Comment

          Working...
          X