Announcement

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

  • Plotting non linearities in regression (as dy/dx) and interaction effects from reghdfe (multiple fixed effects)

    Hello everyone,

    First of all, thank you in advance for any answers to this post. It's very rare to have a forum like this with many serious people sharing knowledge and helping each other out

    FIRST QUESTION:
    I am doing a panel study in which my unit of analysis is a bank, and I follow their net interest margin (NIMQ, dependent variable) over years and quarters. My independent variable is the level of FED interest rate (shortterm), which I also include as a squared term in the model. Another independent variable is ycurve. All the other variables that you see below are controls. Furthermore, I also control for bank and year fixed effects, that is why I am using reghdfe. The regression model is as below:
    Code:
    reghdfe nimq shortterm c.shortterm#c.shortterm ycurve gdpqyoy housing stockmarket hhi laggedatq laggedleverage laggeddeprat laggedcostincrat crisis, a(gvkey qyear) vce(robust)
    For this model, I would like to create a graph similar to the one below, where I plot the effect of a change in the independent variable shortterm on the dependent one. Since there are non-linearities, and the coefficient of shortterm2 is opposite than shortterm, I expect it to be very similar to the one here: https://drive.google.com/file/d/1fSR...ew?usp=sharing.
    I tried by using margins and marginsplot, but it appears that it is not suited for this kind of regression. Does anyone have any suggestions on how I could plot the effect of a change in the independent variable on the dependent (dy/dx) at different levels of the independent variable?

    SECOND QUESTION:
    I expand this model by including the interaction effect of the independent shortterm with deposit ratio deprat and a "post financial crisis" dummy postcrisis. The dependent variable stays the same, and the model is as below:
    Code:
    reghdfe nimq shortterm ycurve deprat c.shortterm#c.deprat c.ycurve#c.deprat c.shortterm#postcrisis c.ycurve#postcrisis c.shortterm#c.deprat#postcrisis c.ycurve#c.deprat#postcrisis gdpqyoy housing stockmarket hhi laggedatq laggedleverage laggeddeprat laggedcostincrat crisis postcrisis, a(gvkey qyear) vce(robust)
    The results for this regression look like this (model 4 in the table): https://drive.google.com/file/d/1mw9...ew?usp=sharing
    As the interpretation of these results is quite cumbersome, I would like to plot the effect of shortterm on the dependent variable at different levels of deprat and with the postcrisis dummy at 0 and 1. If I am not mistaken, I would expect a graph along these lines: https://drive.google.com/file/d/1Xxe...ew?usp=sharing.
    What do you suggest me to do, to have STATA generate a graph for these interactions? As they are complicated even for myself to interpret, I really hope that with your help I will figure this out.

    I hope I've given enough information. In case I didn't, please tell me and I will provide anything ASAP.

    Thank you again!

  • #2
    Hi Pinco,
    What do you mean margins and marginsplot is not suited for this type of regression?
    I run a small example, and seems to work together with reghdfe.
    Code:
    use http://fmwww.bc.edu/RePEc/bocode/o/oaxaca.dta
    reghdfe lnwage exper (c.tenure##c.tenure)#i.female, abs(educ)
    margins, dydx(tenure) at(tenure=(1/10))
    marginsplot
    Fernando

    Comment


    • #3
      Originally posted by FernandoRios View Post
      Hi Pinco,
      What do you mean margins and marginsplot is not suited for this type of regression?
      I run a small example, and seems to work together with reghdfe.
      Code:
      use http://fmwww.bc.edu/RePEc/bocode/o/oaxaca.dta
      reghdfe lnwage exper (c.tenure##c.tenure)#i.female, abs(educ)
      margins, dydx(tenure) at(tenure=(1/10))
      marginsplot
      Fernando
      Hi Fernando,

      After I run the first regression, with fixed effects, I run the following margins command
      Code:
      margins, dydx(shortterm) at(shortterm=(1(1)10))
      and I get this result:
      Code:
      . margins, dydx(shortterm) at(shortterm=(1(1)10))
      
      Average marginal effects                        Number of obs     =     53,345
      Model VCE    : Robust
      
      Expression   : Linear prediction, predict()
      dy/dx w.r.t. : shortterm
      
      1._at        : shortterm       =           1
      
      2._at        : shortterm       =           2
      
      3._at        : shortterm       =           3
      
      4._at        : shortterm       =           4
      
      5._at        : shortterm       =           5
      
      6._at        : shortterm       =           6
      
      7._at        : shortterm       =           7
      
      8._at        : shortterm       =           8
      
      9._at        : shortterm       =           9
      
      10._at       : shortterm       =          10
      
      ------------------------------------------------------------------------------
                   |            Delta-method
                   |      dy/dx   Std. Err.      z    P>|z|     [95% Conf. Interval]
      -------------+----------------------------------------------------------------
      shortterm    |
               _at |
                1  |          .  (not estimable)
                2  |          .  (not estimable)
                3  |          .  (not estimable)
                4  |          .  (not estimable)
                5  |          .  (not estimable)
                6  |          .  (not estimable)
                7  |          .  (not estimable)
                8  |          .  (not estimable)
                9  |          .  (not estimable)
               10  |          .  (not estimable)
      ------------------------------------------------------------------------------
      
      .

      Comment


      • #4
        Oh I just realized, if I run the regression without fixed effects (by using noabsorb instead), margins works, and gives me this result:
        Code:
        . margins, dydx(shortterm) at(shortterm=(1(1)10))
        
        Average marginal effects                        Number of obs     =     53,405
        Model VCE    : Robust
        
        Expression   : Linear prediction, predict()
        dy/dx w.r.t. : shortterm
        
        1._at        : shortterm       =           1
        
        2._at        : shortterm       =           2
        
        3._at        : shortterm       =           3
        
        4._at        : shortterm       =           4
        
        5._at        : shortterm       =           5
        
        6._at        : shortterm       =           6
        
        7._at        : shortterm       =           7
        
        8._at        : shortterm       =           8
        
        9._at        : shortterm       =           9
        
        10._at       : shortterm       =          10
        
        ------------------------------------------------------------------------------
                     |            Delta-method
                     |      dy/dx   Std. Err.      z    P>|z|     [95% Conf. Interval]
        -------------+----------------------------------------------------------------
        shortterm    |
                 _at |
                  1  |   .0986238    .007004    14.08   0.000     .0848962    .1123513
                  2  |   .0760109   .0052646    14.44   0.000     .0656925    .0863293
                  3  |    .053398   .0045439    11.75   0.000     .0444921    .0623039
                  4  |   .0307851   .0052773     5.83   0.000     .0204417    .0411285
                  5  |   .0081722   .0070231     1.16   0.245    -.0055929    .0219373
                  6  |  -.0144407   .0092233    -1.57   0.117    -.0325181    .0036367
                  7  |  -.0370536   .0116227    -3.19   0.001    -.0598336   -.0142735
                  8  |  -.0596665     .01412    -4.23   0.000    -.0873412   -.0319917
                  9  |  -.0822793   .0166714    -4.94   0.000    -.1149546    -.049604
                 10  |  -.1048922   .0192553    -5.45   0.000    -.1426319   -.0671526
        ------------------------------------------------------------------------------
        The relative marginsplot looks similar to what I was expecting, which is reassuring. It also works if I use the bank- fixed effects, but not when I use year- fixed effects (nor both)
        However, do you know why it does not let me calculate the margins when I include the year fixed effects? My panel study is quarter over quarter.
        Also, do you know if the implications of this are "significant", meaning, should I be worried by the fact that margins work only when I don't include year- fixed effects?

        Last question about "SECOND QUESTION" in my original post, how do you suggest me I code the margins command to have a marginsplot result similar to this: https://drive.google.com/file/d/1Xxe...ew?usp=sharing

        Comment


        • #5
          As in Italian Pinco Pallino sounds like John Doe or Mr So-and-So in English, please take a look at FAQ #6. Thanks.
          Kind regards,
          Carlo
          (Stata 19.0)

          Comment


          • #6
            Originally posted by Carlo Lazzaro View Post
            As in Italian Pinco Pallino sounds like John Doe or Mr So-and-So in English, please take a look at FAQ #6. Thanks.
            Ouch... apologies. I must have completely overlooked that. My name is Alfredo Desiderio. I always try to avoid sharing private information online unless necessary, and only now I see that it is necessary here.

            I hope the mods will have mercy on me
            Last edited by Alfredo Desiderio; 15 May 2019, 06:46.

            Comment


            • #7
              Thanks, Alfredo: you can easily change your user name: just click on “Contact us” located at the bottom right-hand corner of every page.
              There's nothing to warn about sharing your given name and family name on this forum. As the FAQ #6 explains, it is only a matter of creating a friendly and professional context among people who might not have ever the chance to meet each other personally.
              If you're afraid about asking/replying in terms that will be judged poor from the scientific board of, say,the Royal Statistical Society (it happens to me quite often, too), just don't mind: all in all particpating to this forum is mainly a matter of continous learning.
              I do hope that you enjoy this forum.
              Last edited by Carlo Lazzaro; 15 May 2019, 07:14.
              Kind regards,
              Carlo
              (Stata 19.0)

              Comment


              • #8
                Originally posted by Carlo Lazzaro View Post
                Thanks, Alfredo: you can easily change your user name: just click on “Contact us” located at the bottom right-hand corner of every page.
                There's nothing to warn about sharing your given name and family name on this forum. As the FAQ #6 explains, it is only a matter of creating a friendly and professional context among people who might not have ever the chance to meet each other personally.
                If you're afraid about asking/replying in terms that will be judged poor from the scientific board of, say,the Royal Statistical Society (it happens to me quite often, too), just don't mind: all in all particpating to this forum is mainly a matter of continous learning.
                I do hope that you enjoy this forum.
                Thank you for the advice, I asked the administrators to change it.
                Have a nice day!
                ---
                As for my original question, the "FIRST QUESTION" has been solved. However, for the "SECOND QUESTION" in the original post, I am having difficulties in getting the "margins" of the IV on the DV for the two levels of the dummy postcrisis, and for varying levels of the deprat continuous variable. I am very new to STATA (I am working on my bachelor's thesis at the moment), so it might be a trivial question for you, but by looking at the STATA documentation and on other posts I cannot find anything suitable (I found many information but only for discrete variables). If anyone has any suggestions on how to have a marginsplot similar to this one I would be infinitely grateful.

                Thank you again.

                Alfredo
                Last edited by Alfredo Desiderio; 15 May 2019, 09:28.

                Comment


                • #9
                  Hi Alfredo,
                  So to my previous example, perhaps this is what you want:

                  Code:
                  use http://fmwww.bc.edu/RePEc/bocode/o/oaxaca.dta, clear
                  reghdfe lnwage exper (c.tenure##c.tenure)#i.female, abs(educ)
                  margins, dydx(tenure) at(tenure=(1/10) female=(0/1))
                  marginsplot

                  Comment


                  • #10
                    Originally posted by FernandoRios View Post
                    Hi Alfredo,
                    So to my previous example, perhaps this is what you want:

                    Code:
                    use http://fmwww.bc.edu/RePEc/bocode/o/oaxaca.dta, clear
                    reghdfe lnwage exper (c.tenure##c.tenure)#i.female, abs(educ)
                    margins, dydx(tenure) at(tenure=(1/10) female=(0/1))
                    marginsplot
                    Thanks for your help!
                    I am using the following regression before the margins command:
                    Code:
                    reghdfe nimq shortterm ycurve deprat c.shortterm#c.deprat c.ycurve#c.deprat c.shortterm#postcrisis c.ycurve#postcrisis c.shortterm#c.deprat#postcrisis c.ycurve#c.deprat#postcrisis gdpqyo
                    > y housing stockmarket hhi laggedatq laggedleverage laggeddeprat laggedcostincrat crisis postcrisis, a(gvkey) vce(robust)
                    Unfortunately when I try this code, I have the error message below:
                    Code:
                    . margins, dydx(deprat) at(deprat=(10(10)100) postcrisis=(0/1))
                    postcrisis ambiguous abbreviation
                    If I then append the "*" after postcrisis I receive this error message:
                    Code:
                    . margins, dydx(deprat) at(deprat=(10(10)100) postcrisis*=(0/1))
                    variable 'postcrisis*' not found in list of covariates
                    If I try to put the "i." before the dummy variables (as you did in your code), I have the following regression results:
                    Code:
                    . reghdfe nimq shortterm ycurve deprat c.shortterm#c.deprat c.ycurve#c.deprat c.shortterm#i
                    > .postcrisis c.ycurve#i.postcrisis c.shortterm#c.deprat#i.postcrisis c.ycurve#c.deprat#i.p
                    > ostcrisis gdpqyoy housing stockmarket hhi laggedatq laggedleverage laggeddeprat laggedcos
                    > tincrat i.crisis i.postcrisis, a(gvkey) vce(robust)
                    (dropped 59 singleton observations)
                    (MWFE estimator converged in 1 iterations)
                    
                    HDFE Linear regression                            Number of obs   =     53,330
                    Absorbing 1 HDFE group                            F(  19,  51539) =     897.06
                                                                      Prob > F        =     0.0000
                                                                      R-squared       =     0.7937
                                                                      Adj R-squared   =     0.7865
                                                                      Within R-sq.    =     0.3445
                                                                      Root MSE        =     0.4173
                    
                    -----------------------------------------------------------------------------------------
                                            |               Robust
                                       nimq |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
                    ------------------------+----------------------------------------------------------------
                                  shortterm |   -.322707   .0233672   -13.81   0.000     -.368507   -.2769071
                                     ycurve |   -.121366    .030827    -3.94   0.000    -.1817873   -.0609448
                                     deprat |  -.0098609   .0023636    -4.17   0.000    -.0144935   -.0052282
                                            |
                       c.shortterm#c.deprat |   .0047284   .0002977    15.88   0.000     .0041449    .0053119
                                            |
                          c.ycurve#c.deprat |   .0018157    .000396     4.59   0.000     .0010396    .0025919
                                            |
                     postcrisis#c.shortterm |
                                         1  |  -2.085626   .6159386    -3.39   0.001    -3.292872   -.8783804
                                            |
                        postcrisis#c.ycurve |
                                         1  |   .4015669   .0625293     6.42   0.000      .279009    .5241249
                                            |
                     postcrisis#c.shortterm#|
                                   c.deprat |
                                         1  |    .032846   .0077025     4.26   0.000      .017749     .047943
                                            |
                        postcrisis#c.ycurve#|
                                   c.deprat |
                                         1  |  -.0041186   .0007803    -5.28   0.000    -.0056481   -.0025892
                                            |
                                    gdpqyoy |   .0144439   .0016711     8.64   0.000     .0111685    .0177192
                                    housing |   -.004207   .0001171   -35.92   0.000    -.0044365   -.0039774
                                stockmarket |  -.0004281   .0000553    -7.75   0.000    -.0005364   -.0003198
                                        hhi |   3.042235   .5290335     5.75   0.000     2.005324    4.079146
                                  laggedatq |  -5.29e-07   4.67e-08   -11.31   0.000    -6.20e-07   -4.37e-07
                             laggedleverage |  -.0377087   .0013181   -28.61   0.000    -.0402922   -.0351252
                               laggeddeprat |   .0133915   .0017108     7.83   0.000     .0100384    .0167446
                           laggedcostincrat |  -.0059021    .000196   -30.11   0.000    -.0062864   -.0055179
                                   1.crisis |  -.1027035   .0101184   -10.15   0.000    -.1225358   -.0828713
                               1.postcrisis |  -.4070331   .0352265   -11.55   0.000    -.4760773   -.3379889
                                      _cons |   7.943936   .1882992    42.19   0.000     7.574868    8.313005
                    -----------------------------------------------------------------------------------------
                    
                    Absorbed degrees of freedom:
                    -----------------------------------------------------+
                     Absorbed FE | Categories  - Redundant  = Num. Coefs |
                    -------------+---------------------------------------|
                           gvkey |      1772           0        1772     |
                    -----------------------------------------------------+
                    So I try to execute margins again and it works, but the results are not exactly what i am looking for. As far as I see, they don't register the effect of shortterm on the dependent variable at different levels of deprat and postcrisis:
                    Code:
                    . margins, dydx(deprat) at(deprat=(10(10)100) postcrisis=(0/1))
                    
                    Average marginal effects                        Number of obs     =     53,330
                    Model VCE    : Robust
                    
                    Expression   : Linear prediction, predict()
                    dy/dx w.r.t. : deprat
                    
                    1._at        : deprat          =          10
                                   postcrisis      =           0
                    
                    2._at        : deprat          =          10
                                   postcrisis      =           1
                    
                    3._at        : deprat          =          20
                                   postcrisis      =           0
                    
                    4._at        : deprat          =          20
                                   postcrisis      =           1
                    
                    5._at        : deprat          =          30
                                   postcrisis      =           0
                    
                    6._at        : deprat          =          30
                                   postcrisis      =           1
                    
                    7._at        : deprat          =          40
                                   postcrisis      =           0
                    
                    8._at        : deprat          =          40
                                   postcrisis      =           1
                    
                    9._at        : deprat          =          50
                                   postcrisis      =           0
                    
                    10._at       : deprat          =          50
                                   postcrisis      =           1
                    
                    11._at       : deprat          =          60
                                   postcrisis      =           0
                    
                    12._at       : deprat          =          60
                                   postcrisis      =           1
                    
                    13._at       : deprat          =          70
                                   postcrisis      =           0
                    
                    14._at       : deprat          =          70
                                   postcrisis      =           1
                    
                    15._at       : deprat          =          80
                                   postcrisis      =           0
                    
                    16._at       : deprat          =          80
                                   postcrisis      =           1
                    
                    17._at       : deprat          =          90
                                   postcrisis      =           0
                    
                    18._at       : deprat          =          90
                                   postcrisis      =           1
                    
                    19._at       : deprat          =         100
                                   postcrisis      =           0
                    
                    20._at       : deprat          =         100
                                   postcrisis      =           1
                    
                    ------------------------------------------------------------------------------
                                 |            Delta-method
                                 |      dy/dx   Std. Err.      z    P>|z|     [95% Conf. Interval]
                    -------------+----------------------------------------------------------------
                    deprat       |
                             _at |
                              1  |   .0070919   .0017531     4.05   0.000     .0036559    .0105279
                              2  |   .1016581   .0228552     4.45   0.000     .0568628    .1464534
                              3  |   .0070919   .0017531     4.05   0.000     .0036559    .0105279
                              4  |   .1016581   .0228552     4.45   0.000     .0568628    .1464534
                              5  |   .0070919   .0017531     4.05   0.000     .0036559    .0105279
                              6  |   .1016581   .0228552     4.45   0.000     .0568628    .1464534
                              7  |   .0070919   .0017531     4.05   0.000     .0036559    .0105279
                              8  |   .1016581   .0228552     4.45   0.000     .0568628    .1464534
                              9  |   .0070919   .0017531     4.05   0.000     .0036559    .0105279
                             10  |   .1016581   .0228552     4.45   0.000     .0568628    .1464534
                             11  |   .0070919   .0017531     4.05   0.000     .0036559    .0105279
                             12  |   .1016581   .0228552     4.45   0.000     .0568628    .1464534
                             13  |   .0070919   .0017531     4.05   0.000     .0036559    .0105279
                             14  |   .1016581   .0228552     4.45   0.000     .0568628    .1464534
                             15  |   .0070919   .0017531     4.05   0.000     .0036559    .0105279
                             16  |   .1016581   .0228552     4.45   0.000     .0568628    .1464534
                             17  |   .0070919   .0017531     4.05   0.000     .0036559    .0105279
                             18  |   .1016581   .0228552     4.45   0.000     .0568628    .1464534
                             19  |   .0070919   .0017531     4.05   0.000     .0036559    .0105279
                             20  |   .1016581   .0228552     4.45   0.000     .0568628    .1464534
                    ------------------------------------------------------------------------------
                    What am I doing wrong here.

                    Again, infinite thanks for your help!

                    Have a good day/evening!

                    Alfredo

                    Comment


                    • #11
                      Actually, one difference between my regression and yours (@Fernando) I've noticed (and might be the reason of the problem), is that yours shows the coefficients of the interaction between tenure and female also when female is 0.
                      Code:
                      . reghdfe lnwage exper (c.tenure##c.tenure)#i.female, abs(educ)
                      (MWFE estimator converged in 1 iterations)
                      
                      HDFE Linear regression                            Number of obs   =      1,434
                      Absorbing 1 HDFE group                            F(   5,   1419) =      45.69
                                                                        Prob > F        =     0.0000
                                                                        R-squared       =     0.2915
                                                                        Adj R-squared   =     0.2845
                                                                        Within R-sq.    =     0.1387
                                                                        Root MSE        =     0.4492
                      
                      ------------------------------------------------------------------------------------------
                                        lnwage |      Coef.   Std. Err.      t    P>|t|     [95% Conf. Interval]
                      -------------------------+----------------------------------------------------------------
                                         exper |   .0102495   .0014964     6.85   0.000     .0073141    .0131849
                                               |
                               female#c.tenure |
                                            0  |    .037078   .0046988     7.89   0.000     .0278606    .0462954
                                            1  |   .0274781   .0055165     4.98   0.000     .0166568    .0382995
                                               |
                      female#c.tenure#c.tenure |
                                            0  |  -.0009028    .000154    -5.86   0.000     -.001205   -.0006007
                                            1  |  -.0009372   .0002225    -4.21   0.000    -.0013737   -.0005007
                                               |
                                         _cons |   3.078262   .0236702   130.05   0.000     3.031829    3.124694
                      ------------------------------------------------------------------------------------------
                      
                      Absorbed degrees of freedom:
                      -----------------------------------------------------+
                       Absorbed FE | Categories  - Redundant  = Num. Coefs |
                      -------------+---------------------------------------|
                              educ |        10           0          10     |
                      -----------------------------------------------------+
                      In my regression, the lowcrisis=0 appears to be the "baseline", and so it does not show a coefficient for it.

                      Might this be the cause?

                      Comment


                      • #12
                        Edit:
                        Apparently I was making some syntax errors.
                        By using your code I solved the SECOND QUESTION as well.
                        Code:
                        . margins, dydx(shortterm) at(deprat=(10(10)100) postcrisis=(0 1))
                        
                        Average marginal effects                        Number of obs     =     53,330
                        Model VCE    : Robust
                        
                        Expression   : Linear prediction, predict()
                        dy/dx w.r.t. : shortterm
                        
                        1._at        : deprat          =          10
                                       postcrisis      =           0
                        
                        2._at        : deprat          =          10
                                       postcrisis      =           1
                        
                        3._at        : deprat          =          20
                                       postcrisis      =           0
                        
                        4._at        : deprat          =          20
                                       postcrisis      =           1
                        
                        5._at        : deprat          =          30
                                       postcrisis      =           0
                        
                        6._at        : deprat          =          30
                                       postcrisis      =           1
                        
                        7._at        : deprat          =          40
                                       postcrisis      =           0
                        
                        8._at        : deprat          =          40
                                       postcrisis      =           1
                        
                        9._at        : deprat          =          50
                                       postcrisis      =           0
                        
                        10._at       : deprat          =          50
                                       postcrisis      =           1
                        
                        11._at       : deprat          =          60
                                       postcrisis      =           0
                        
                        12._at       : deprat          =          60
                                       postcrisis      =           1
                        
                        13._at       : deprat          =          70
                                       postcrisis      =           0
                        
                        14._at       : deprat          =          70
                                       postcrisis      =           1
                        
                        15._at       : deprat          =          80
                                       postcrisis      =           0
                        
                        16._at       : deprat          =          80
                                       postcrisis      =           1
                        
                        17._at       : deprat          =          90
                                       postcrisis      =           0
                        
                        18._at       : deprat          =          90
                                       postcrisis      =           1
                        
                        19._at       : deprat          =         100
                                       postcrisis      =           0
                        
                        20._at       : deprat          =         100
                                       postcrisis      =           1
                        
                        ------------------------------------------------------------------------------
                                     |            Delta-method
                                     |      dy/dx   Std. Err.      z    P>|z|     [95% Conf. Interval]
                        -------------+----------------------------------------------------------------
                        shortterm    |
                                 _at |
                                  1  |  -.2754233   .0204318   -13.48   0.000    -.3154689   -.2353777
                                  2  |   -2.03259   .5424151    -3.75   0.000    -3.095704   -.9694756
                                  3  |  -.2281395   .0175104   -13.03   0.000    -.2624593   -.1938197
                                  4  |  -1.656846   .4655017    -3.56   0.000    -2.569212   -.7444792
                                  5  |  -.1808558   .0146115   -12.38   0.000    -.2094938   -.1522178
                                  6  |  -1.281102   .3888146    -3.29   0.001    -2.043165   -.5190394
                                  7  |   -.133572   .0117516   -11.37   0.000    -.1566048   -.1105393
                                  8  |  -.9053583   .3125202    -2.90   0.004    -1.517887     -.29283
                                  9  |  -.0862883   .0089682    -9.62   0.000    -.1038657   -.0687108
                                 10  |  -.5296145   .2369981    -2.23   0.025    -.9941222   -.0651069
                                 11  |  -.0390045   .0063626    -6.13   0.000    -.0514749   -.0265341
                                 12  |  -.1538708   .1633231    -0.94   0.346    -.4739782    .1662366
                                 13  |   .0082792   .0042731     1.94   0.053    -.0000958    .0166543
                                 14  |    .221873   .0958537     2.31   0.021     .0340032    .4097427
                                 15  |    .055563   .0037094    14.98   0.000     .0482926    .0628334
                                 16  |   .5976167    .060898     9.81   0.000     .4782588    .7169747
                                 17  |   .1028468   .0051947    19.80   0.000     .0926653    .1130282
                                 18  |   .9733605   .1011749     9.62   0.000     .7750614     1.17166
                                 19  |   .1501305   .0076114    19.72   0.000     .1352124    .1650487
                                 20  |   1.349104    .169621     7.95   0.000     1.016653    1.681555
                        ------------------------------------------------------------------------------
                        And the plot is very similar to what I was looking for:
                        Click image for larger version

Name:	graph 1c deprat postcrisis.png
Views:	1
Size:	51.5 KB
ID:	1498568


                        Thank you very much!

                        Happy STATA to all and I hope to come back to this forum in the future to contribute instead of asking

                        Comment

                        Working...
                        X