Announcement

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

  • Dummy variables in xtpmg model

    Dear all,
    I have a dynamic panel data (N=35 T=24 not balanced).
    After months of work the -xtpmg- model finally converges (all dfe, mg, and pmg options). When I tried to include dummy variable (for the SR eq.) as in the code that follows
    Code:
    xi: xtpmg d.socialspending d(1).gin  d(1).gdp d(1).democracy_ind (1).sixty_five_l i.latin_america, lr(l.socialspending gini gdp democracy_ind)  replace ec(ec) pmg
    Yet, it omits the dummy variable i.latin_america despite the fact that 1/4 of the sample answers the dummy criterion. The same holds for other dummies I have tried to include in the model.
    I wonder what might be the reason for that.
    Thanks,
    Anat



  • #2
    Knowing nothing whatever about the technique used by Pooled Mean Group estimation (xtpmg is apparently a user-written program from SSC), I'll hazard a guess that the problem is that the indicator variables you are trying to include are constant within each country over time (countries don't join or leave Latin America over time), while the PMG technique implicitly or explicitly includes a fixed effect for each country in your data. This is just a guess, you need to go back to the literature on PMG to see if this is indeed the case.

    Comment


    • #3
      This makes perfectly sense! I will re-read the pertaining literature and, if indeed its not possible, I will split the sample or use interaction variable. Thank you very much!

      Comment


      • #4
        Dear Anatmanes, I have similar problem please .. could you tell me how you handled

        Comment


        • #5
          Dear Mazen Ahmad,
          Eventually what I did was to run the xtpmg on sub-groups of the entire sample, I have found that Samargandi et al. (2015) in their paper in World Development have actually did the same. Though you have to have enough countries in each sub-sample you run I guess.
          Hope it helps,
          Anat

          Comment


          • #6
            Dear all I am running this command for PMG
            xtpmg d.lnbd d.lnds d.lngdp elec l.elec f.elec, lr(l.lnbd lnds lngdp elec l.elec f.elec ) ec(ecm) replace pm Election year is my dummy variable i want to check impact of election year, pre-election year and post-election year on Budget deficit.
            1) Can i put d. (dot) with elecion year dummy in short run?
            2) Can we check dummy for long run also, or we should not include it in long run?
            if i use d.(dot) d.elec d.l.elec d.f.elec in short run. Then lag l.elec is omitted from long run. without use of d.(dot) i got all dummies results sig and in right direction. is this result correct without use of d.(dot).


            what is the write pattern?
            and lastly how to interpret result as dep var is in log foam while dummy elec is without log.
            Thanking you



            Comment


            • #7
              I can also answer partially, I think technically you can take a first difference of the var. election, but obviously election is a stationary variable, so there is no point in doing so. I think that all variables should appear in both long and short time specification..

              Comment


              • #8
                i also tried this in Eviews 9 using pmg. i put election year dummies in fixed regressors the output shows dummies in short run without D() and it does not include dummies in long run. so this means that we should not put d. in short run in stata command while in longrun we should not include dummies considering them as fix regressors .(according to output of eviews 9). do you think i am right?

                Comment


                • #9
                  I have the same problem in using dummy variable in stata Dr Muhammad Nauman
                  Did you reach its solution as I searched but I got no answer
                  Thanks a lot

                  Comment


                  • #10
                    HANAN ABOELFARAG what exactly is your problem, Muhammad Nauman had a very specific problem.

                    Comment


                    • #11
                      I have a trend variable t and a dummy variable for political instability
                      I want to put them in pmg model
                      I want to ask whether to put them both in short run (with difference d.) and also in the long run or not???
                      thanks a lot

                      Comment


                      • #12
                        Hanan, the variable for political stability you can enter to both long and short run specifications, I do not see why not. Regarding trend variable, I'm not sure why you need such a variable in an error-correction models. I haven't tackled a research that employed ECM with time trends (but I may be wrong). I think that if you want to account for external shocks (like financial crisis or a change in the political regime) you can test for structural breaks using -clemao1 and clemio1 - which allow for a single structural break. But I think you should have long enough time dimension in order to be able to test for structural breaks.

                        Comment


                        • #13
                          Guys, I have gone with an idea and solved it with my problem. The below is the code of my model and I have a dummy variable named 'dumfc'. I use this proxy without difference command (d.) at short-run command box and it treats my dummy variable as fixed regressor which we can do easily at Eviews. For the note: I use data from 1985 to 2016, my dummy variable 'dumfc' is the dummy for financial crises of 1997 onwards as represented by binary number 1 till 2016 to see the fixed long-run effect of financial crises. My model is ARDL (2, 1, 1,1, 1).
                          xtpmg d(1/2).lngdppc d.lndcp d.lntrd d.lngov d.lninf dumfc, lr(l.lngdppc l.lndcp l.lntrd l.lngov l.lninf) replace ec(ECT) pmg

                          If I am right please feedback me, if I am wrong then please, suggest me. Let's learn together.

                          Comment

                          Working...
                          X