Announcement

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

  • FE Logit model does not converge, whereas RE probit does, misspecification?

    I have given Panel data and want to investigate on the determinants of collateral incidence, binary variable.
    input float Collateraldummy byte Numberofemployees long Totalassets float Corporationdummy long Grossprofit double(Profitability Leverage) long Loansize byte Maturity str14 Industry double(GDPGrowth Duration) byte Housebank str6 Loantype float Banks
    1 28 1500 1 1600 .0625 .95 475 10 "Other Industry" .015 0 0 "Credit" 7
    0 28 1500 1 1600 .0625 .95 475 10 "Other Industry" .015 0 0 "Credit" 1
    1 15 500 1 800 .0875 .5 150 10 "Other Industry" .015 5.75 1 "Credit" 8
    1 15 500 1 800 .0875 .5 30 1 "Other Industry" .015 5.75 1 "LC" 8
    1 15 500 1 800 .0875 .5 20 1 "Other Industry" .015 6 1 "LC" 8
    1 10 387 0 815 .0343558282208589 .72 80 1 "Handcraft" .022 10 1 "LC" 8
    1 10 415 0 830 .05060240963855422 .77 80 1 "Handcraft" .022 11 1 "LC" 8
    1 10 400 0 850 .03529411764705882 .9 120 1 "Handcraft" .015 12 1 "LC" 8
    0 10 415 0 830 .05060240963855422 .77 60 6 "Handcraft" .022 1 0 "Credit" 7
    1 25 800 1 3500 .03428571428571429 .2 100 1 "Other Industry" .015 4.666666666666667 0 "LC" 3
    1 25 800 1 3500 .03428571428571429 .2 620 20 "Other Industry" .015 0 0 "Credit" 6
    1 25 800 1 3500 .03428571428571429 .2 230 3 "Other Industry" .015 5 0 "LC" 5
    0 8 130 0 300 .23333333333333334 .4 50 10 "Gastronomic" .015 4.75 1 "Credit" 1
    0 3 60 0 190 0 0 20 10 "Gastronomic" .005 0 1 "Credit" 1
    0 8 130 0 300 .23333333333333334 .4 15 3 "Gastronomic" .022 3 0 "LC" 3
    1 12 450 1 800 .08125 .26 50 10 "Handcraft" .015 10.083333333333334 0 "Credit" 8
    1 12 462 1 830 .0819277108433735 .32 125 5 "Handcraft" .022 8 0 "Credit" 8
    1 12 438 1 755 .07549668874172186 .3 100 5 "Handcraft" .022 0 0 "Credit" 4
    1 12 450 1 800 .08125 .26 15 1 "Handcraft" .015 10 0 "LC" 8
    1 12 438 1 755 .07549668874172186 .3 15 1 "Handcraft" .022 9 0 "LC" 8
    1 12 462 1 830 .0819277108433735 .32 15 1 "Handcraft" .022 8 0 "LC" 8
    1 12 438 1 755 .07549668874172186 .3 120 1 "Handcraft" .022 10 0 "LC" 5
    1 12 462 1 830 .0819277108433735 .32 120 1 "Handcraft" .022 9 0 "LC" 5
    0 12 450 1 800 .08125 .26 10 1 "Handcraft" .015 10.583333333333334 0 "LC" 5
    1 10 320 1 1000 .08 .55 70 6 "Gastronomic" .015 7 0 "Credit" 5
    1 10 320 1 1000 .08 .55 100 5 "Gastronomic" .015 5.166666666666667 0 "Credit" 4
    1 12 277 1 800 .09375 .6 150 4 "Gastronomic" .015 5.083333333333333 1 "Credit" 5
    1 25 720 1 1800 .11388888888888889 .45 350 3 "Gastronomic" .022 12 1 "Credit" 5
    0 25 695 1 2000 .105 .45 300 6 "Gastronomic" .015 14 1 "Credit" 5
    1 3 248 1 500 .11 .44 30 4 "Handcraft" .017 0 0 "Credit" 3
    1 3 250 1 600 .08333333333333333 .5 50 5 "Handcraft" .022 1.33 0 "Credit" 4
    0 3 248 1 500 .11 .44 8 1 "Handcraft" .017 0 0 "LC" 7
    0 3 250 1 600 .08333333333333333 .5 8 1 "Handcraft" .022 1 0 "LC" 7
    0 3 250 1 600 .08333333333333333 .5 10 3 "Handcraft" .022 1.083 0 "LC" 9
    1 25 462 1 1750 .022857142857142857 .45 100 1 "Handcraft" .022 0 0 "LC" 9
    1 29 450 1 1900 .027105263157894736 .5 200 3 "Handcraft" .022 .5833333333333334 0 "LC" 9
    1 29 450 1 1900 .027105263157894736 .5 100 1 "Handcraft" .022 1 0 "LC" 9
    1 25 462 1 1750 .022857142857142857 .45 250 5 "Handcraft" .022 0 0 "Credit" 4
    1 29 440 1 2000 .025 .5 200 5 "Handcraft" .015 1.4166666666666667 0 "Credit" 4
    1 9 360 1 415 .18795180722891566 .25 15 1 "Handcraft" .017 5 1 "LC" 7
    1 9 350 1 435 .18620689655172415 .25 25 1 "Handcraft" .022 6 1 "LC" 7
    1 9 345 1 430 .18604651162790697 .3 15 1 "Handcraft" .022 7 1 "LC" 7
    1 14 1000 0 1450 .07931034482758621 .6 350 7 "Gastronomic" .005 15 1 "Credit" 7
    0 15 1050 0 1500 .06666666666666667 .7 300 10 "Gastronomic" .015 20 1 "Credit" 7
    1 14 1000 0 1450 .07931034482758621 .6 150 1 "Gastronomic" .005 15 1 "LC" 7
    1 15 970 0 1400 .06785714285714285 .7 150 1 "Gastronomic" .022 16.5 1 "LC" 7
    1 15 960 0 1475 .06779661016949153 .7 150 1 "Gastronomic" .017 17.75 1 "LC" 7
    1 3 350 0 400 .125 .5 20 1 "Handcraft" .015 7 1 "LC" 6
    1 3 350 0 400 .125 .5 15 5 "Handcraft" .015 7 1 "Credit" 6
    0 25 500 1 1100 .18181818181818182 .8 150 10 "Handcraft" .015 15 1 "Credit" 5
    0 25 500 1 1100 .18181818181818182 .8 400 15 "Handcraft" .015 15 1 "Credit" 5
    0 25 500 1 1100 .18181818181818182 .8 50 1 "Handcraft" .015 15 1 "LC" 5
    0 25 620 0 2000 .15 .2 150 10 "Handcraft" .015 20 1 "Credit" 5
    0 25 620 0 2000 .15 .2 50 1 "Handcraft" .015 20 1 "LC" 5
    0 12 380 1 1500 .06666666666666667 .3 25 5 "Handcraft" .015 15 1 "Credit" 5
    1 7 400 0 950 .1368421052631579 .25 300 5 "Handcraft" .017 3 1 "Credit" 7
    0 9 425 0 1000 .123 .2 250 7 "Handcraft" .015 6 1 "Credit" 7
    1 7 400 0 950 .1368421052631579 .25 50 1 "Handcraft" .017 3 1 "LC" 7
    1 8 415 0 975 .14358974358974358 .2 80 1 "Handcraft" .022 4.333333333333333 1 "LC" 7
    1 9 410 0 935 .13368983957219252 .2 80 1 "Handcraft" .022 5.333333333333333 1 "LC" 7
    1 9 425 0 1000 .123 .2 80 1 "Handcraft" .015 6 1 "LC" 7
    1 6 370 0 427 .14285714285714285 .42 80 5 "Handcraft" .022 23 1 "Credit" 5
    1 6 370 0 427 .14285714285714285 .42 30 1 "Handcraft" .022 8 0 "LC" 6
    1 6 375 0 430 .13953488372093023 .45 45 1 "Handcraft" .022 8.75 0 "LC" 6
    0 6 370 0 427 .14285714285714285 .42 80 5 "Handcraft" .022 0 0 "Credit" 2
    0 28 3500 1 2875 .05495652173913043 .38 500 10 "Other Industry" .005 14 1 "Credit" 5
    0 30 3625 1 3000 .05 .4 400 7 "Other Industry" .022 4 0 "Credit" 3
    1 30 3625 1 3000 .05 .4 60 2 "Other Industry" .022 5 0 "LC" 4
    1 30 3625 1 3000 .05 .4 50 2 "Other Industry" .022 .16666666666666666 0 "LC" 4
    0 15 3100 1 2600 .06538461538461539 .5 150 3 "Other Industry" .018 5 0 "Credit" 3
    0 15 3100 1 2600 .06538461538461539 .5 130 4 "Other Industry" .018 4 0 "Credit" 4
    0 15 3100 1 2600 .06538461538461539 .5 50 2 "Other Industry" .018 4 0 "LC" 4
    1 35 2650 1 2300 .09 .21 300 5 "Other Industry" .022 22 1 "Credit" 7
    1 35 2710 1 2425 .09278350515463918 .28 250 7 "Other Industry" .017 23 1 "Credit" 7
    0 33 2665 1 2400 .0875 .25 50 9 "Other Industry" .022 25.25 1 "Credit" 7
    0 33 2700 1 2350 .08297872340425531 .25 80 10 "Other Industry" .015 26.333333333333332 1 "Credit" 7
    1 34 2710 1 2425 .09278350515463918 .28 80 1 "Other Industry" .017 23.166666666666668 1 "LC" 7
    1 26 1980 1 1650 .0893939393939394 .26 325 10 "Handcraft" .022 16 1 "Credit" 7
    0 26 2050 1 1700 .08941176470588236 .31 150 8 "Handcraft" .022 18.333333333333332 1 "Credit" 7
    0 26 1930 1 1750 .08857142857142856 .33 220 5 "Handcraft" .015 19.166666666666668 1 "Credit" 7
    0 26 2050 1 1700 .08941176470588236 .31 80 1 "Handcraft" .022 18.166666666666668 1 "LC" 7
    end
    [/CODE]
    I ran the following models:
    Code:
    xtset Banks
    Code:
    xtprobit Collateraldummy  Numberofemployees Totalassets Corporationdummy Grossprofit Profitability Leverage Loansize Maturity g1 g3 GDPGrowth Duration Housebank if Loantype=="Crédit"
    Code:
    xtlogit Collateraldummy  Numberofemployees Totalassets Corporationdummy Grossprofit Profitability Leverage Loansize Maturity g1 g3 GDPGrowth Duration Housebank if Loantype=="Credit", fe
    and the latter model does not converge. Can someone assist me why fixed effects model does not converge?

  • #2
    Marcel:
    provided that I fail to get why you performed an -xtprobit, re- (and not an -xtlogit, re-) if you want to compare it with -xtlogit, fe-, the reason might be due to the conditional MLE that Stata uses for conditional fixed effect speciifcation (due to the incidental parameter bias).
    That's all I can say, considering that, being inconsistent with the codes you ran, your data excerpt does not allow any further try out:
    Code:
    . xtprobit Collateraldummy  Numberofemployees Totalassets Corporationdummy Grossprofit Profitability Leverage Loansize M
    > aturity g1 g3 GDPGrowth Duration Housebank if Loantype=="Crédit"
    variable g1 not found
    r(111);
    
    . xtlogit Collateraldummy  Numberofemployees Totalassets Corporationdummy Grossprofit Profitability Leverage Loansize Ma
    > turity g1 g3 GDPGrowth Duration Housebank if Loantype=="Credit", fe
    variable g1 not found
    r(111);
    Kind regards,
    Carlo
    (Stata 19.0)

    Comment


    • #3
      Hi Carlo, I forgot to add:
      Code:
      tabulate Industry, generate (g)
      Also by running xtlogit fe, the model Fails to converge.

      Comment


      • #4
        Marcel:
        this is what I got when I tried out your codes on the data excerpt that you provided:
        Code:
        . xtlogit Collateraldummy  Numberofemployees Totalassets Corporationdummy Grossprofit Profitability Leverage Loansize Ma
        > turity g1 g3 GDPGrowth Duration Housebank if Loantype=="Crédit"
        number of quadrature points must be less than or equal to number of obs
        r(198);
        
        . xtlogit Collateraldummy  Numberofemployees Totalassets Corporationdummy Grossprofit Profitability Leverage Loansize Ma
        > turity g1 g3 GDPGrowth Duration Housebank if Loantype=="Crédit", fe
        no observations
        r(2000);
        Kind regards,
        Carlo
        (Stata 19.0)

        Comment


        • #5
          Carlo: You mistyped credit in the loan type condition by puting an Accent on the e.
          Code:
          xtlogit Collateraldummy Numberofemployees Totalassets Corporationdummy Grossprofit Profitability Leverage Loansize Maturity g1 g3 GDPGrowth Duration Housebank if Loantype=="Credit"
          With this command it should work

          Comment


          • #6
            Marcel:
            Code:
            . xtlogit Collateraldummy  Numberofemployees Totalassets Corporationdummy Grossprofit Profitability Leverage Loansize Ma
            > turity g1 g3 GDPGrowth Duration Housebank if Loantype=="Credit"
            
            Fitting comparison model:
            
            Iteration 0:   log likelihood = -28.406838 
            Iteration 1:   log likelihood = -15.561767 
            Iteration 2:   log likelihood = -15.054675 
            Iteration 3:   log likelihood =  -15.03877 
            Iteration 4:   log likelihood = -15.038719 
            Iteration 5:   log likelihood = -15.038719 
            
            Fitting full model:
            
            tau =  0.0     log likelihood = -15.038719
            tau =  0.1     log likelihood = -15.028571
            tau =  0.2     log likelihood = -15.020202
            tau =  0.3     log likelihood = -15.014881
            tau =  0.4     log likelihood = -15.015193
            
            Iteration 0:   log likelihood = -15.014881  (not concave)
            Iteration 1:   log likelihood = -14.938414  (not concave)
            Iteration 2:   log likelihood = -14.701337  (not concave)
            Iteration 3:   log likelihood = -14.495394 
            Iteration 4:   log likelihood = -14.468065  (not concave)
            Iteration 5:   log likelihood = -14.456505  (not concave)
            Iteration 6:   log likelihood = -14.143409  (not concave)
            Iteration 7:   log likelihood = -13.405297 
            Iteration 8:   log likelihood = -12.959901  (not concave)
            Iteration 9:   log likelihood = -12.308177  (not concave)
            Iteration 10:  log likelihood = -12.271018  (not concave)
            Iteration 11:  log likelihood = -12.225136  (not concave)
            Iteration 12:  log likelihood = -12.217646  (not concave)
            Iteration 13:  log likelihood = -12.046979  (not concave)
            Iteration 14:  log likelihood = -12.041297  (not concave)
            Iteration 15:  log likelihood = -12.018284  (not concave)
            Iteration 16:  log likelihood =  -12.01632  (not concave)
            Iteration 17:  log likelihood = -12.016161  (not concave)
            Iteration 18:  log likelihood = -12.016161  (not concave)
            Iteration 19:  log likelihood = -11.986078  (not concave)
            Iteration 20:  log likelihood = -11.922856  (not concave)
            Iteration 21:  log likelihood = -11.856966 
            Iteration 22:  log likelihood = -11.360595  (not concave)
            Iteration 23:  log likelihood = -11.251977  (not concave)
            Iteration 24:  log likelihood = -11.197457  (not concave)
            Iteration 25:  log likelihood = -11.179545  (not concave)
            Iteration 26:  log likelihood = -11.165182  (not concave)
            Iteration 27:  log likelihood = -11.154235  (not concave)
            Iteration 28:  log likelihood = -11.145972  (not concave)
            Iteration 29:  log likelihood = -11.138874  (not concave)
            Iteration 30:  log likelihood = -11.133031 
            Iteration 31:  log likelihood = -11.099705  (not concave)
            Iteration 32:  log likelihood = -11.082318  (not concave)
            Iteration 33:  log likelihood = -11.068342  (not concave)
            Iteration 34:  log likelihood = -11.059078  (not concave)
            Iteration 35:  log likelihood = -11.052853  (not concave)
            Iteration 36:  log likelihood =   -11.0434  (not concave)
            Iteration 37:  log likelihood = -11.021579 
            Iteration 38:  log likelihood = -10.979063 
            Iteration 39:  log likelihood = -10.918318  (not concave)
            Iteration 40:  log likelihood = -10.895011 
            Iteration 41:  log likelihood = -10.865415 
            Iteration 42:  log likelihood = -10.857044 
            Iteration 43:  log likelihood = -10.857001 
            Iteration 44:  log likelihood = -10.857001 
            
            Random-effects logistic regression              Number of obs     =         41
            Group variable: Banks                           Number of groups  =          8
            
            Random effects u_i ~ Gaussian                   Obs per group:
                                                                          min =          1
                                                                          avg =        5.1
                                                                          max =         13
            
            Integration method: mvaghermite                 Integration pts.  =         12
            
                                                            Wald chi2(13)     =      40.74
            Log likelihood  = -10.857001                    Prob > chi2       =     0.0001
            
            -----------------------------------------------------------------------------------
              Collateraldummy |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
            ------------------+----------------------------------------------------------------
            Numberofemployees |  -.2238301   .5919023    -0.38   0.705    -1.383937    .9362771
                  Totalassets |  -.0100239   .0041471    -2.42   0.016     -.018152   -.0018958
             Corporationdummy |   20.49138   8.722236     2.35   0.019     3.396111    37.58665
                  Grossprofit |  -.0325263    .007744    -4.20   0.000    -.0477042   -.0173484
                Profitability |   -167.662   107.0115    -1.57   0.117    -377.4007    42.07678
                     Leverage |  -33.90284    12.5921    -2.69   0.007    -58.58291   -9.222771
                     Loansize |   .1152236   .0388577     2.97   0.003     .0390638    .1913834
                     Maturity |  -2.170337   1.209309    -1.79   0.073     -4.54054    .1998656
                           g1 |   8.950802    5.89968     1.52   0.129    -2.612358    20.51396
                           g3 |   29.32534   7.670234     3.82   0.000     14.29195    44.35872
                    GDPGrowth |    318.578     460.42     0.69   0.489    -583.8286    1220.984
                     Duration |   1.576345   .6858259     2.30   0.022     .2321506    2.920539
                    Housebank |  -13.14557   8.036533    -1.64   0.102    -28.89688    2.605748
                        _cons |   47.79136   19.64879     2.43   0.015     9.280439    86.30229
            ------------------+----------------------------------------------------------------
                     /lnsig2u |   5.656499    .686119                      4.311731    7.001268
            ------------------+----------------------------------------------------------------
                      sigma_u |   16.91583   5.803135                       8.63536    33.13645
                          rho |   .9886335   .0077101                       .957746    .9970128
            -----------------------------------------------------------------------------------
            LR test of rho=0: chibar2(01) = 8.36                   Prob >= chibar2 = 0.002
            Thanks for the fix; -re- gives the results reported above; -fe- does not converge.
            It may be that -fe- is not the way to go, as it unfeasible with your data.
            Kind regards,
            Carlo
            (Stata 19.0)

            Comment


            • #7
              Thanks Carlo, is there any way I can investigate on the reason why fixed effects does not converge?

              Comment


              • #8
                Marcel:
                I would recommend to scrutinize your dataset and investigate whether a constant (or a predictor with scant variation) is the culprit.
                In this instance, the usual recipe is to start it all over again, adding one predictor at time and see when convergence problems begin.
                Kind regards,
                Carlo
                (Stata 19.0)

                Comment


                • #9
                  Thx again Carlo! I encountered also a Problem when running a random effects model of the "LC" data:
                  Code:
                  xtlogit Collateraldummy Age Numberofemployees Totalassets Corporationdummy Profitability Leverage Loansize Maturity Duration Housebank GDPGrowth if Loantype=="LC",re vce(cluster Banks)
                  The coeeficient and the Standard error for profitability and for gdpgrowth are exorbitantly high. I checked for multicollinearity and found only for Grossprofit a vif of above 6 so I ommitted it
                  Code:
                  regress Collateraldummy Age Numberofemployees Totalassets Corporationdummy Grossprofit Profitability g1 g3 GDPGrowth Leverage Leverage Loansize Maturity Duration Housebank if Loantype=="LC", vce(cluster Banks)
                  Code:
                  estat vif
                  Profitability and gdpgrowth have a Pretty low Variation across my data and I Question whether to include them or not? Can someone assist me pls?

                  Comment


                  • #10
                    Marcel:
                    while is probably wise to omit predictor that show high correlation with other regressors (although this in not always the way to go: linear and squared terms of the same predictor are highly correlated but shoud be kept together notwithstanding in the right-hand side of the regression equation if you there0evidence of turning point) the issue concerning which predictors should be included depends on the data generating process, not on your data.
                    Kind regards,
                    Carlo
                    (Stata 19.0)

                    Comment


                    • #11
                      Thx again Carlo! Could you please tell me what test I could run in order to choose between a pooled logit clustered by Banks and the logit random effects Modell? Many thanks in Advance.

                      Comment


                      • #12
                        Marcel:
                        as far as I know, there's no such a test.
                        You may want to compare -xtlogit, pa- with -corr(exchangeable)- option vs -xtlogit,re-, both with -vce(bootstrap) standard errors (200 bootstrap replications are recommended; however, in the following toy-example for sake of brevity only 50 bootstrap replications have been performed):
                        Code:
                        use http://www.stata-press.com/data/r15/union
                        . xtlogit union age grade not_smsa south##c.year, pa corr(exch) vce(robust)
                        
                        Iteration 1: tolerance = .14878775
                        Iteration 2: tolerance = .00949339
                        Iteration 3: tolerance = .00040606
                        Iteration 4: tolerance = .00001602
                        Iteration 5: tolerance = 6.628e-07
                        
                        GEE population-averaged model                   Number of obs     =     26,200
                        Group variable:                     idcode      Number of groups  =      4,434
                        Link:                                logit      Obs per group:
                        Family:                           binomial                    min =          1
                        Correlation:                  exchangeable                    avg =        5.9
                                                                                      max =         12
                                                                        Wald chi2(6)      =     154.88
                        Scale parameter:                         1      Prob > chi2       =     0.0000
                        
                                                         (Std. Err. adjusted for clustering on idcode)
                        ------------------------------------------------------------------------------
                                     |               Robust
                               union |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
                        -------------+----------------------------------------------------------------
                                 age |   .0165893    .008951     1.85   0.064    -.0009543    .0341329
                               grade |   .0600669   .0133193     4.51   0.000     .0339616    .0861722
                            not_smsa |  -.1215445   .0613803    -1.98   0.048    -.2418477   -.0012412
                             1.south |  -1.857094   .5389238    -3.45   0.001    -2.913366   -.8008231
                                year |  -.0121168   .0096998    -1.25   0.212    -.0311282    .0068945
                                     |
                        south#c.year |
                                  1  |   .0160193   .0067217     2.38   0.017      .002845    .0291937
                                     |
                               _cons |   -1.39755   .5603767    -2.49   0.013    -2.495868   -.2992317
                        ------------------------------------------------------------------------------
                        
                        . xtlogit union age grade not_smsa south##c.year, pa corr(exch) vce(bootstrap)
                        (running xtlogit on estimation sample)
                        
                        Bootstrap replications (50)
                        ----+--- 1 ---+--- 2 ---+--- 3 ---+--- 4 ---+--- 5
                        ..................................................    50
                        
                        GEE population-averaged model                   Number of obs     =     26,200
                        Group variable:                     idcode      Number of groups  =      4,434
                        Link:                                logit      Obs per group:
                        Family:                           binomial                    min =          1
                        Correlation:                  exchangeable                    avg =        5.9
                                                                                      max =         12
                                                                        Wald chi2(6)      =     200.57
                        Scale parameter:                         1      Prob > chi2       =     0.0000
                        
                                                      (Replications based on 4,434 clusters in idcode)
                        ------------------------------------------------------------------------------
                                     |   Observed   Bootstrap                         Normal-based
                               union |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
                        -------------+----------------------------------------------------------------
                                 age |   .0165893   .0089015     1.86   0.062    -.0008573    .0340359
                               grade |   .0600669   .0148367     4.05   0.000     .0309874    .0891464
                            not_smsa |  -.1215445   .0708016    -1.72   0.086    -.2603131    .0172241
                             1.south |  -1.857094   .5321013    -3.49   0.000    -2.899994    -.814195
                                year |  -.0121168   .0094942    -1.28   0.202    -.0307251    .0064914
                                     |
                        south#c.year |
                                  1  |   .0160193   .0065568     2.44   0.015     .0031682    .0288705
                                     |
                               _cons |   -1.39755   .5838289    -2.39   0.017    -2.541833   -.2532663
                        ------------------------------------------------------------------------------
                        
                        . xtlogit union age grade not_smsa south##c.year, re  vce(bootstrap)
                        (running xtlogit on estimation sample)
                        
                        Bootstrap replications (50)
                        ----+--- 1 ---+--- 2 ---+--- 3 ---+--- 4 ---+--- 5
                        ..................................................    50
                        
                        Random-effects logistic regression              Number of obs     =     26,200
                        Group variable: idcode                          Number of groups  =      4,434
                        
                        Random effects u_i ~ Gaussian                   Obs per group:
                                                                                      min =          1
                                                                                      avg =        5.9
                                                                                      max =         12
                        
                        Integration method: mvaghermite                 Integration pts.  =         12
                        
                                                                        Wald chi2(6)      =     188.57
                        Log likelihood  = -10540.274                    Prob > chi2       =     0.0000
                        
                                                      (Replications based on 4,434 clusters in idcode)
                        ------------------------------------------------------------------------------
                                     |   Observed   Bootstrap                         Normal-based
                               union |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
                        -------------+----------------------------------------------------------------
                                 age |   .0156732   .0117927     1.33   0.184    -.0074402    .0387865
                               grade |   .0870851   .0181596     4.80   0.000      .051493    .1226773
                            not_smsa |  -.2511884   .0981912    -2.56   0.011    -.4436396   -.0587372
                             1.south |  -2.839112   .9119219    -3.11   0.002    -4.626446   -1.051778
                                year |  -.0068604   .0147007    -0.47   0.641    -.0356733    .0219524
                                     |
                        south#c.year |
                                  1  |   .0238506    .011305     2.11   0.035     .0016932    .0460079
                                     |
                               _cons |  -3.009365   .8467094    -3.55   0.000    -4.668885   -1.349845
                        -------------+----------------------------------------------------------------
                            /lnsig2u |   1.749366   .0475537                      1.656163     1.84257
                        -------------+----------------------------------------------------------------
                             sigma_u |   2.398116   .0570196                      2.288923    2.512517
                                 rho |   .6361098   .0110074                      .6142744    .6573985
                        ------------------------------------------------------------------------------
                        LR test of rho=0: chibar2(01) = 6004.43                Prob >= chibar2 = 0.000
                        
                        .
                        In this case I would go -re-.
                        Kind regards,
                        Carlo
                        (Stata 19.0)

                        Comment


                        • #13
                          Is there any reason why stata does not report the LR test in re Output?

                          xtlogit Collateraldummy Age Numberofemployees Totalassets Corporationdummy Pr
                          > ofitability Leverage Loansize Maturity Duration Housebank GDPGrowth if Loanty
                          > pe=="LC",re vce(cluster Banks)

                          Fitting comparison model:

                          Iteration 0: log pseudolikelihood = -21.326554
                          Iteration 1: log pseudolikelihood = -11.42658
                          Iteration 2: log pseudolikelihood = -9.5766355
                          Iteration 3: log pseudolikelihood = -8.9169841
                          Iteration 4: log pseudolikelihood = -8.8011437
                          Iteration 5: log pseudolikelihood = -8.7958644
                          Iteration 6: log pseudolikelihood = -8.7958433
                          Iteration 7: log pseudolikelihood = -8.7958433

                          Fitting full model:

                          tau = 0.0 log pseudolikelihood = -8.7958433
                          tau = 0.1 log pseudolikelihood = -8.8177919

                          Iteration 0: log pseudolikelihood = -8.8177919
                          Iteration 1: log pseudolikelihood = -8.8009727
                          Iteration 2: log pseudolikelihood = -8.7971467
                          Iteration 3: log pseudolikelihood = -8.7961565
                          Iteration 4: log pseudolikelihood = -8.795914
                          Iteration 5: log pseudolikelihood = -8.7958576
                          Iteration 6: log pseudolikelihood = -8.7958456
                          Iteration 7: log pseudolikelihood = -8.7958435
                          Iteration 8: log pseudolikelihood = -8.7958435

                          Calculating robust standard errors:

                          Random-effects logistic regression Number of obs = 40
                          Group variable: Banks Number of groups = 7

                          Random effects u_i ~ Gaussian Obs per group:
                          min = 2
                          avg = 5.7
                          max = 14

                          Integration method: mvaghermite Integration pts. = 12

                          Wald chi2(6) = .
                          Log pseudolikelihood = -8.7958435 Prob > chi2 = .

                          (Std. Err. adjusted for 7 clusters in Banks)
                          ------------------------------------------------------------------------------
                          | Robust
                          Collateral~y | Coef. Std. Err. z P>|z| [95% Conf. Interval]
                          -------------+----------------------------------------------------------------
                          Age | .2842625 .0934242 3.04 0.002 .1011545 .4673706
                          Numberofem~s | -.0557256 .1385207 -0.40 0.687 -.3272211 .2157699
                          Totalassets | -.0006608 .0008383 -0.79 0.431 -.0023039 .0009822
                          Corporatio~y | 3.118555 1.891975 1.65 0.099 -.5896478 6.826758
                          Profitabil~y | -116.2015 83.85258 -1.39 0.166 -280.5496 48.14651
                          Leverage | -24.51085 17.22751 -1.42 0.155 -58.27614 9.254439
                          Loansize | .0293048 .0190263 1.54 0.124 -.0079859 .0665956
                          Maturity | -2.993109 3.169984 -0.94 0.345 -9.206164 3.219946
                          Duration | -1.075623 .5260166 -2.04 0.041 -2.106597 -.0446497
                          Housebank | 13.65171 6.676909 2.04 0.041 .5652081 26.73821
                          GDPGrowth | -51.92392 168.8189 -0.31 0.758 -382.803 278.9551
                          _cons | 22.72995 17.88084 1.27 0.204 -12.31585 57.77575
                          -------------+----------------------------------------------------------------
                          /lnsig2u | -13.68694 . . .
                          -------------+----------------------------------------------------------------
                          sigma_u | .0010664 . . .
                          rho | 3.46e-07 . . .
                          ------------------------------------------------------------------------------


                          Comment


                          • #14
                            Marcel:
                            yes, and it's due to non-default standard errors.
                            Kind regards,
                            Carlo
                            (Stata 19.0)

                            Comment


                            • #15
                              Thx again Carlo. If I cannot reject the null hypothesis of the LR test that the Panel estimator is different from the pooled estimator, can I then choose between the re and pooled model or is the pooled preferred?

                              Comment

                              Working...
                              X