Announcement

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

  • How to find the Ttest of comparing between two coefficients

    Hello everyone,
    I have used two independent vars in my paper ( cfo_turnover, ceo_turnover)
    I have used each one individually in one model, but I want to compare between the two coefficients of all variables and find the Ttest . Where it will be in a column.
    please what is the command?
    or how i can find it?
    kind regards

  • #2
    your question is not completely clear to me, but I think this is where to start:
    Code:
    help suest

    Comment


    • #3
      Originally posted by Rich Goldstein View Post
      your question is not completely clear to me, but I think this is where to start:
      Code:
      help suest
      Thanks for replying
      I will attach a picture of Table that explains what I want to ask.
      in the model 3 of the table, you will see results of comparing two coefficients for each variable.
      how can I run this model?
      Click image for larger version

Name:	C9BA12D9-5DA0-4BA3-86F2-CB066725CA15.jpeg
Views:	2
Size:	256.2 KB
ID:	1541229

      Comment


      • #4
        Fernando Martins please I couldn’t find the column ( differences). I need to find the Ttest value for each variable.

        Comment


        • #5
          Hi Alkebsee,

          If I got it right, you have two models that you estimated and want to compare if the estimated coefficients are statistically different between models, right?

          What kind of models are you using? They are panel data models or cross section ones?

          As Rich suggested one option is to use the 'suest' command, but this does not work with panel data regressions (xtreg). So, depending on what kind of data you have, the approach is different.

          Comment


          • #6
            Originally posted by Fernando Martins View Post
            Hi Alkebsee,

            If I got it right, you have two models that you estimated and want to compare if the estimated coefficients are statistically different between models, right?
            Yes, that is what I want exactly.


            What kind of models are you using? They are panel data models or cross section ones?
            I am using a cross-section one.

            As Rich suggested one option is to use the 'suest' command, but this does not work with panel data regressions (xtreg). So, depending on what kind of data you have, the approach is different.
            actually I have used the suest command, but this command gives me only one value ( coefficient and p-value). While in the picture above they provided a value for each variable.
            is that possible with the suest command ? If yes so I think I don’t know how to apply it.
            please how to extract a value for each variable?
            Fernando Martins thank you so much

            Comment


            • #7
              Well, if 'suest' works for you, then you can regress the first model, store its results, run the second model, store its results, then compare them with the 'test' command. This will yield a chi2 statistic with 1 degree of freedom. You can run this for each variable. Basically you would be comparing if the coefficients are the same in each model or not.

              You can do as follows:
              Code:
              regress y1 x1 x2 x3
              estimates store model1
              
              regress y2 x1 x2 x3
              estimates store model2
              
              suest model1 model2
              
              test [model1_mean]x1 = [model2_mean]x1
              This should yield something like I got here in an example:
              Code:
               test [model1_mean]lfert=[model2_mean]lfert
              
               ( 1)  [model1_mean]lfert - [model2_mean]lfert = 0
              
                         chi2(  1) =    7.37
                       Prob > chi2 =    0.0066
              The coefficient of the variable 'lfert' is the same (at 1% significance) for both models. In other words, the difference between the coefficients is statistically indifferent from zero at 1% significance.

              Comment


              • #8
                Originally posted by Fernando Martins View Post
                Well, if 'suest' works for you, then you can regress the first model, store its results, run the second model, store its results, then compare them with the 'test' command. This will yield a chi2 statistic with 1 degree of freedom. You can run this for each variable. Basically you would be comparing if the coefficients are the same in each model or not.

                You can do as follows:
                Code:
                regress y1 x1 x2 x3
                estimates store model1
                
                regress y2 x1 x2 x3
                estimates store model2
                
                suest model1 model2
                
                test [model1_mean]x1 = [model2_mean]x1
                As i have noticed that the Variable X1 ( which is the independent variable) is the same variable in to models, right?
                while in my models the independent variable is not the same . Which means I want to compare between the effects of two different independent variables ( CEO’s and CFO’s pay) on the dependent variable ( earnings management). It’s the same process?
                or there’s another way to do it?
                one more thing
                if I want to get the differences between the two models for each variable ( x1,2,3, etc). How to get this thing done?

                This should yield something like I got here in an example:
                Code:
                test [model1_mean]lfert=[model2_mean]lfert
                
                ( 1) [model1_mean]lfert - [model2_mean]lfert = 0
                
                chi2( 1) = 7.37
                Prob > chi2 = 0.0066
                Thanks for this explanation, but I understood that the beta you have shown ( chi2 and its P value) means there’s no significant difference, right?
                if yes, so I have to get p value > 0.10 to say there’s a significant difference between the two models, right?

                The coefficient of the variable 'lfert' is the same (at 1% significance) for both models. In other words, the difference between the coefficients is statistically indifferent from zero at 1% significance.
                Thank you so much.

                Comment


                • #9
                  Fernando Martins

                  Comment

                  Working...
                  X