Hi all,

I'm interested in running a probability model but I have some questions regarding with the models itself and the control variables. Before the questions about the model, here's a section of my data set:

Code:
1996    1    1    -.3275632
1997    1    1    -.37988619
1998    1    1    -.35953836
1999    1    0    -.38094193
2000    1    1    -.2752674
2001    1    0    -.3891454
2002    1    1    -.4167883
2003    1    1    -.18402586
2004    1    1    -.32373861
2005    1    0    -.26459956
2006    1    0    -.24347964
2007    1    1    -.22100002
2008    1    1    -.30709973
2009    1    0    -.29539555
2010    1    0    -.39313385
2011    1    0    -.36766358
2012    1    0    -.37702698
2013    1    0    -.44015804
2014    1    0    -.24885774
2015    1    0    -.31083205
2016    1    0    -.27822521
2017    1    0    -.22154936
1995    2    0    .07744394
1996    2    1    .03747911
1997    2    0    -.13627645
1998    2    0    .15022897
1999    2    0    .05101566
2000    2    0    .05925428
Please note that I deliberately excluded other control variables.

First, second, third, and fourth columns show the years, companies, binary dependent variable, and the continuous independent variable (leveragedeficit in the following tables), respectively.

With a data set like that, I need to use -xtprobit- or -xtlogit- instead of only -probit- or -logit-, right?

I followed following steps:

Code:
. xtset CompanyNo BaseYear
       panel variable:  CompanyNo (unbalanced)
        time variable:  BaseYear, 1995 to 2017, but with gaps
                delta:  1 unit
Code:
 xtprobit AcquirerStatus LeverageDeficit Size Markettobook Profitability i.BaseYear

Fitting comparison model:

Iteration 0:   log likelihood = -703.83892  
Iteration 1:   log likelihood = -668.72656  
Iteration 2:   log likelihood = -667.89193  
Iteration 3:   log likelihood = -667.88916  
Iteration 4:   log likelihood = -667.88916  

Fitting full model:

rho =  0.0     log likelihood = -667.88916
rho =  0.1     log likelihood = -659.08637
rho =  0.2     log likelihood = -660.71391

Iteration 0:   log likelihood = -659.08647  
Iteration 1:   log likelihood = -656.04543  
Iteration 2:   log likelihood =  -655.8358  
Iteration 3:   log likelihood = -655.83545  
Iteration 4:   log likelihood = -655.83545  

Random-effects probit regression                Number of obs     =      2,554
Group variable: CompanyNo                       Number of groups  =        248

Random effects u_i ~ Gaussian                   Obs per group:
                                                              min =          1
                                                              avg =       10.3
                                                              max =         23

Integration method: mvaghermite                 Integration pts.  =         12

                                                Wald chi2(26)     =      42.88
Log likelihood  = -655.83545                    Prob > chi2       =     0.0199

---------------------------------------------------------------------------------
 AcquirerStatus |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
----------------+----------------------------------------------------------------
LeverageDeficit |  -.4894872   .2504744    -1.95   0.051     -.980408    .0014337
           Size |   .2125652   .0846811     2.51   0.012     .0465932    .3785371
   Markettobook |   .0508144   .0591066     0.86   0.390    -.0650325    .1666613
  Profitability |   .3614916   .5259731     0.69   0.492    -.6693968     1.39238
                |
       BaseYear |
          1996  |  -.0196722   .3486411    -0.06   0.955    -.7029962    .6636518
          1997  |  -.1248034   .3488386    -0.36   0.721    -.8085145    .5589076
          1998  |   .0540659   .3247021     0.17   0.868    -.5823385    .6904703
          1999  |   .3028502   .3145904     0.96   0.336    -.3137356     .919436
          2000  |  -.4904255   .3670268    -1.34   0.181    -1.209785    .2289339
          2001  |  -.1871212   .3393259    -0.55   0.581    -.8521878    .4779454
          2002  |  -.1044416   .3337692    -0.31   0.754    -.7586173    .5497341
          2003  |   .0790474   .3210148     0.25   0.805      -.55013    .7082248
          2004  |  -.1749998   .3335653    -0.52   0.600    -.8287758    .4787761
          2005  |  -.1767613   .3258052    -0.54   0.587    -.8153278    .4618051
          2006  |  -.1462115   .3139424    -0.47   0.641    -.7615273    .4691043
          2007  |   .0661997   .3027836     0.22   0.827    -.5272452    .6596446
          2008  |  -.7384657   .3527952    -2.09   0.036    -1.429932   -.0469998
          2009  |  -.0535747   .3010514    -0.18   0.859    -.6436246    .5364752
          2010  |  -.3910161   .3195727    -1.22   0.221    -1.017367    .2353349
          2011  |  -.0690389   .2992925    -0.23   0.818    -.6556415    .5175637
          2012  |   -.485205   .3220254    -1.51   0.132    -1.116363    .1459532
          2013  |  -.3406325   .3115639    -1.09   0.274    -.9512865    .2700215
          2014  |  -.3151666    .310039    -1.02   0.309    -.9228319    .2924987
          2015  |  -.1495515   .3018954    -0.50   0.620    -.7412556    .4421526
          2016  |  -.0080481    .300264    -0.03   0.979    -.5965547    .5804585
          2017  |  -.5602969   .3589999    -1.56   0.119    -1.263924    .1433299
                |
          _cons |  -3.327422    .783768    -4.25   0.000    -4.863579   -1.791265
----------------+----------------------------------------------------------------
       /lnsig2u |  -1.586501   .3264732                     -2.226377   -.9466255
----------------+----------------------------------------------------------------
        sigma_u |   .4523719   .0738437                      .3285099    .6229352
            rho |   .1698767   .0460388                      .0974067     .279564
---------------------------------------------------------------------------------
LR test of rho=0: chibar2(01) = 24.11                  Prob >= chibar2 = 0.000
I know fixed effects do not work with -xtprobit- command. However, I added time dummies by typing i.BaseYear. Is this the correct way to control year effects? I guess Stata runs a random effects regression under -xtprobit- command by default, right? Do insignificant year dummies mean something specifically? Can -xtlogit- be more advantageous over -xtprobit- when I want to control year effects?

As you can see from the result table, I don't have a Pseudo R2. Is it because of something I did wrong?

Finally, I wanted to see the marginal effects and typed

Code:
margins LeverageDeficit
and I was given the following:

Code:
LeverageDeficit:  factor variables may not contain noninteger values
r(452);
even if the the variable is not a factor variable. What could be the reason of this?

Sorry for the long post and lots of questions.

Thanks!