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

  • F-test for difference between two coefficients

    Dear all,

    how do you interpret the output of
    test b = c

    I regress
    reg a b c
    and then type the above test. The output comes in the form

    . test active = inactive

    ( 1) active - inactive = 0

    F( 1, 11420) = 5.78
    Prob > F = 0.0162

    I believe that if (Prob > F) <= 0.05, the difference is significant, as in that the coefficients to b and c are statistically different from each other. Is this correct?

    I know this is a basic question, but I want to make sure because some results I am getting do not match my intuition.

  • #2
    your interpretation is correct.
    As an aside. if some results sound as unexpected when you contrast them to theoretical assumptions, I would take a look at your regression model specification (-regres postestimation- included).
    Kind regards,
    (Stata 15.1 SE)


    • #3
      Keep in mind that you have a pretty large N, which makes it pretty easy for even small effects and differences to be statistically significant. You'll have to judge for yourself how substantively significant you feel the differences are. You might also consider using the .01 level of significance given that you have a fairly large sample.
      Richard Williams, Notre Dame Dept of Sociology
      Stata Version: 15.1MP (2 processor)

      EMAIL: rwilliam@ND.Edu


      • #4
        Unnecessary statement.


        • #5
          I didn't recognize what my post might look like. I was correcting a unnecessary statement I made, not suggesting the other statements were unnecessary. Of course Carlo and Richard's statements are clear and to the point.


          • #6
            obviously no problem.
            Rather,a wise warning about any unnecessary statement I made/will make in my posts (LOL)!I
            Kind regards,
            (Stata 15.1 SE)


            • #7
              Thanks everyone, this forum is awesome.

              Is it possible to obtain the output from a post regression command like this somehow?

              Regardless of this, I could really need some help to interpret the output. I should know this, but here are two rather stupid questions from me:

              I'm also running a test to compare two coefficients in one regression.
              In F tests we usually talk about F(df1, df2) referring to the critical values for given degrees of freedom in two samples. In my case, I'm looking at F(1,2397). What does that really mean?
              And how do i interpret the number after Prob > F?

              Would be very happy if anyone could help!