Best model for this data?

[Sue Bowers]

Dear all,
I'm having slight problems finding the right model for my data. My independent variable is an ordered categorical variable, scale 0-3. My dependent variables are summarized below:

    Variable |       Obs        Mean    Std. Dev.       Min        Max
-------------+--------------------------------------------------------
     PFbuyin |      1370    .2734774    .2943603          0       .992
csrcountr~og |      1370   -.6122415    .2217881  -1.203973  -.3147107
indivle~tlog |      1370    1.764782    .8460997          0   2.822278
 orglegitlog |      1370    3.291117    .5551468          0   3.882576
leftvotes100 |      1370    11.27494    18.62507          0       51.2
-------------+--------------------------------------------------------
greenvot~100 |      1370    3.612737    3.725471          0       11.8
performori~v |      1370    5.929653    .2981624       5.09       6.41
diversity~20 |       865    .2947977    .4562155          0          1
diversity2~3 |       865    .3063584    .4612471          0          1
diversityo~3 |       865    .1560694    .3631311          0          1
-------------+--------------------------------------------------------
aum_natura~g |      1370    2.947121     1.86649          0   8.177789
   instcollv |      1370    4.475164    .4076155       3.84       5.57
    turnover |      1370    .2734128    .1520528          0      .8425
ftse4goode~r |      1370    .0510949    .2202717          0          1

The dataset is a panel dataset. The dependent variable is time-variant. The indepvars are a mixture of time variant and time invariant. They are however mostly time invariant. So I would guess that in this case, I should perform a cross section analysis with an ologit or mlogit? But when I try, I get the below results:

. bysort Datayear: ologit  percentintESG PFbuyin csrcountryrating_log  indivlegitlog orglegitlog leftvotes100 greenvotes100 performori
> entv diversityupto20 diversity20to33 diversityover33  aum_naturallog instcollv turnover   ftse4goodever if IM==1, nolog

--------------------------------------------------------------------------------------------------------------------------------------
-> Datayear = 2007
no observations

--------------------------------------------------------------------------------------------------------------------------------------
-> Datayear = 2008
note: ftse4goodever omitted because of collinearity
convergence not achieved

Ordered logistic regression                       Number of obs   =         14
                                                  LR chi2(-6)     =      64.53
                                                  Prob > chi2     =          .
Log likelihood =          0                       Pseudo R2       =     1.0000

--------------------------------------------------------------------------------------
       percentintESG |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
---------------------+----------------------------------------------------------------
             PFbuyin |   5974.335          .        .       .            .           .
csrcountryrating_log |  -2804.866          .        .       .            .           .
       indivlegitlog |  -490.9564          .        .       .            .           .
         orglegitlog |   9944.791          .        .       .            .           .
        leftvotes100 |  -89.48098   1.12e+08    -0.00   1.000    -2.20e+08    2.20e+08
       greenvotes100 |   359.1942   5.42e+08     0.00   1.000    -1.06e+09    1.06e+09
      performorientv |  -3898.506   7.56e+08    -0.00   1.000    -1.48e+09    1.48e+09
     diversityupto20 |  -2467.232          .        .       .            .           .
     diversity20to33 |  -1835.009          .        .       .            .           .
     diversityover33 |    319.082          .        .       .            .           .
      aum_naturallog |   146.1296   1.09e+09     0.00   1.000    -2.14e+09    2.14e+09
           instcollv |  -2872.946          .        .       .            .           .
            turnover |  -7486.932          .        .       .            .           .
       ftse4goodever |          0  (omitted)
---------------------+----------------------------------------------------------------
               /cut1 |  -2529.599          .                             .           .
               /cut2 |  -2437.345          .                             .           .
               /cut3 |   -2350.64          .                             .           .
               /cut4 |  -2262.752          .                             .           .
               /cut5 |  -2175.402          .                             .           .
               /cut6 |  -2088.195          .                             .           .
               /cut7 |  -2001.205          .                             .           .
               /cut8 |  -1914.021          .                             .           .
               /cut9 |  -1826.072          .                             .           .
              /cut10 |  -1736.209          .                             .           .
--------------------------------------------------------------------------------------
Note: 14 observations completely determined.  Standard errors questionable.
convergence not achieved
r(430);

. bysort Datayear: mlogit  percentintESG PFbuyin csrcountryrating_log  indivlegitlog orglegitlog leftvotes100 greenvotes100 performori
> entv diversityupto20 diversity20to33 diversityover33  aum_naturallog instcollv turnover   ftse4goodever if IM==1, nolog

--------------------------------------------------------------------------------------------------------------------------------------
-> Datayear = 2007
no observations

--------------------------------------------------------------------------------------------------------------------------------------
-> Datayear = 2008
note: ftse4goodever omitted because of collinearity
invalid matrix stripe;
.18
r(198);

.

Is there something I'm missing with the data characteristics or is there an alternative model I could use? When I just perform a logistic regression on the data, the results are good and normal. But with most time-invariant variables, these results may be spurious, am I right?
Many thanks in advance for your input.

Sue

[Carlo Lazzaro]

Sue:
have you given it a try with -xtologit-?

[Joseph Coveney]

It looks as if it's something that you haven't noticed in the data characteristics.

First, when IM==1, you have zero all-nonmissing observations for one of the two years and fourteen all-nonmissing observations for the other.

Second, your response variable is supposed to be ordered-categorical with a 0-to-3 scale, but there are 10 cut points (11 categories) in the one ordered-logistic model where you had any observations at all.

You might want to look into some data-exploration commands that Stata has, such as inspect, codebook with its options, and the misstable suite.

[Sue Bowers]

Thank you so much for the replies!
Carlo, yes I have tried xtlogit, as that's the model that I'm using for the other binary outcome in this same study. I'm going to paste the results at the end of this post, they are good. However, logit is made for binary outcomes, not for ordered categorical outcomes, isn't it? For that reason I thought logit was not appropriate here.

Joseph:
The data is quite sparse in many years unfortunately.
however, the bigger issue here is that it was 11PM in my time zone when I was writing the post and I accidentally used the wrong outcome variable! Which is continuous 0-100 and so of course it has 10 cut points...

So here is the result of the same model with the actual 0-3 scale:

. bysort Datayear: ologit   seekESGinf PFbuyin csrcountryrating_log  indivlegitlog orglegitlog leftvotes100 gre
> envotes100 performorientv diversityupto20 diversity20to33 diversityover33  aum_naturallog instcollv turnover 
>   ftse4goodever if IM==1, nolog

---------------------------------------------------------------------------------------------------------------
-> Datayear = 2007
note: leftvotes100 omitted because of collinearity
note: performorientv omitted because of collinearity
note: diversity20to33 omitted because of collinearity
note: diversityover33 omitted because of collinearity
note: aum_naturallog omitted because of collinearity
note: instcollv omitted because of collinearity
note: turnover omitted because of collinearity
note: ftse4goodever omitted because of collinearity
convergence not achieved

Ordered logistic regression                       Number of obs   =          7
                                                  LR chi2(-2)     =      14.06
                                                  Prob > chi2     =          .
Log likelihood =          0                       Pseudo R2       =     1.0000

--------------------------------------------------------------------------------------
          seekESGinf |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
---------------------+----------------------------------------------------------------
             PFbuyin |   56858.87          .        .       .            .           .
csrcountryrating_log |  -246340.6          .        .       .            .           .
       indivlegitlog |  -8457.285          .        .       .            .           .
         orglegitlog |   5000.852          .        .       .            .           .
        leftvotes100 |          0  (omitted)
       greenvotes100 |  -558.9055          .        .       .            .           .
      performorientv |          0  (omitted)
     diversityupto20 |   7.27e-09          .        .       .            .           .
     diversity20to33 |          0  (omitted)
     diversityover33 |          0  (omitted)
      aum_naturallog |          0  (omitted)
           instcollv |          0  (omitted)
            turnover |          0  (omitted)
       ftse4goodever |          0  (omitted)
---------------------+----------------------------------------------------------------
               /cut1 |   131653.6          .                             .           .
               /cut2 |   132055.3          .                             .           .
--------------------------------------------------------------------------------------
Note: 7 observations completely determined.  Standard errors questionable.
convergence not achieved
r(430);

Here is the -mlogit- results

 mlogit   seekESGinf PFbuyin csrcountryrating_log  indivlegitlog orglegitlog leftvotes100 greenvotes100 perfor
> morientv diversityupto20 diversity20to33 diversityover33  aum_naturallog instcollv turnover   ftse4goodever i
> f IM==1, nolog

Multinomial logistic regression                   Number of obs   =        244
                                                  LR chi2(56)     =      86.63
                                                  Prob > chi2     =     0.0054
Log likelihood = -317.02112                       Pseudo R2       =     0.1202

--------------------------------------------------------------------------------------
          seekESGinf |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
---------------------+----------------------------------------------------------------
0                    |
             PFbuyin |   3.770004   1.619984     2.33   0.020     .5948926    6.945115
csrcountryrating_log |   1.533198   1.938039     0.79   0.429    -2.265288    5.331684
       indivlegitlog |   .8190287   .5811609     1.41   0.159    -.3200257    1.958083
         orglegitlog |   1.531857   1.189603     1.29   0.198     -.799722    3.863435
        leftvotes100 |    .051257   .0221307     2.32   0.021     .0078817    .0946323
       greenvotes100 |  -.0609428    .104654    -0.58   0.560     -.266061    .1441753
      performorientv |  -.4703705   1.334249    -0.35   0.724     -3.08545    2.144709
     diversityupto20 |  -.5276525   .6308494    -0.84   0.403    -1.764095    .7087895
     diversity20to33 |  -.1591482   .6245496    -0.25   0.799    -1.383243    1.064947
     diversityover33 |    .052211   .8105047     0.06   0.949    -1.536349    1.640771
      aum_naturallog |   -.373716   .1220578    -3.06   0.002    -.6129448   -.1344872
           instcollv |  -2.305154   1.341631    -1.72   0.086    -4.934702    .3243942
            turnover |    .076484   1.250001     0.06   0.951    -2.373472     2.52644
       ftse4goodever |   .0027051   .9701154     0.00   0.998    -1.898686    1.904096
               _cons |   6.487465   6.911163     0.94   0.348    -7.058166     20.0331
---------------------+----------------------------------------------------------------
1                    |
             PFbuyin |  -1.527266   1.292944    -1.18   0.238    -4.061389    1.006857
csrcountryrating_log |   3.621053   2.319897     1.56   0.119     -.925861    8.167968
       indivlegitlog |   .3712943   .5106464     0.73   0.467    -.6295542    1.372143
         orglegitlog |   .2065105     .55732     0.37   0.711    -.8858167    1.298838
        leftvotes100 |   .0313474   .0250906     1.25   0.212    -.0178293    .0805241
       greenvotes100 |   .0013825   .1060251     0.01   0.990    -.2064228    .2091879
      performorientv |  -.5549323   1.561133    -0.36   0.722    -3.614697    2.504832
     diversityupto20 |  -.3730413    .640901    -0.58   0.561    -1.629184    .8831015
     diversity20to33 |  -.4602533   .6426445    -0.72   0.474    -1.719813    .7993068
     diversityover33 |    .392721   .8731194     0.45   0.653    -1.318562    2.104004
      aum_naturallog |  -.3174318   .1234826    -2.57   0.010    -.5594532   -.0754103
           instcollv |   1.314182    1.23309     1.07   0.287    -1.102629    3.730994
            turnover |  -1.164575   1.347346    -0.86   0.387    -3.805324    1.476175
       ftse4goodever |   .8610953   .8458731     1.02   0.309    -.7967854    2.518976
               _cons |  -.9187421   7.499439    -0.12   0.902    -15.61737    13.77989
---------------------+----------------------------------------------------------------
2                    |
             PFbuyin |  -2.631532   1.293166    -2.03   0.042    -5.166091   -.0969737
csrcountryrating_log |  -1.133705   1.614273    -0.70   0.482    -4.297622    2.030213
       indivlegitlog |   .1851066   .4373599     0.42   0.672    -.6721031    1.042316
         orglegitlog |   .4990353   .5279203     0.95   0.345    -.5356696     1.53374
        leftvotes100 |   .0156402   .0197184     0.79   0.428    -.0230072    .0542876
       greenvotes100 |  -.0122351   .0872856    -0.14   0.889    -.1833117    .1588415
      performorientv |  -2.310456   1.378648    -1.68   0.094    -5.012558    .3916449
     diversityupto20 |  -.5146675   .5460846    -0.94   0.346    -1.584974    .5556387
     diversity20to33 |  -.7893414   .5574105    -1.42   0.157    -1.881846     .303163
     diversityover33 |  -.8875393   .7787749    -1.14   0.254     -2.41391    .6388315
      aum_naturallog |  -.0552637     .09708    -0.57   0.569     -.245537    .1350096
           instcollv |   2.642445   1.222674     2.16   0.031     .2460486    5.038842
            turnover |   .3848736   1.108225     0.35   0.728    -1.787207    2.556954
       ftse4goodever |   .5751586   .7674675     0.75   0.454      -.92905    2.079367
               _cons |   .2373194   6.675694     0.04   0.972     -12.8468    13.32144
---------------------+----------------------------------------------------------------
3                    |  (base outcome)
---------------------+----------------------------------------------------------------
4                    |
             PFbuyin |  -1.455721   3.459283    -0.42   0.674    -8.235791    5.324349
csrcountryrating_log |   .7154868   3.578684     0.20   0.842    -6.298604    7.729578
       indivlegitlog |   1.078545   1.595024     0.68   0.499    -2.047645    4.204735
         orglegitlog |   2.559555   3.599231     0.71   0.477    -4.494809    9.613919
        leftvotes100 |   .0799398   .0577578     1.38   0.166    -.0332634    .1931429
       greenvotes100 |  -.3356311   .3140651    -1.07   0.285    -.9511873    .2799251
      performorientv |  -1.369141   2.949892    -0.46   0.643    -7.150823    4.412542
     diversityupto20 |   -1.14568    .985871    -1.16   0.245    -3.077952    .7865913
     diversity20to33 |  -.7222027   .9125344    -0.79   0.429    -2.510737    1.066332
     diversityover33 |   -1.26377   1.375179    -0.92   0.358    -3.959072    1.431531
      aum_naturallog |    .096329   .1774614     0.54   0.587     -.251489     .444147
           instcollv |   3.490562   3.457829     1.01   0.313    -3.286659    10.26778
            turnover |   1.920136    2.18872     0.88   0.380    -2.369677    6.209949
       ftse4goodever |  -13.93619   798.7745    -0.02   0.986    -1579.506    1551.633
               _cons |  -19.96759   19.85417    -1.01   0.315    -58.88104    18.94585
--------------------------------------------------------------------------------------

And here are the results of the panel logistic on the same data:

xtlogit   seekESGinf PFbuyin csrcountryrating_log  indivlegitlog orglegitlog leftvotes100 greenvotes100 perfo
> rmorientv diversityupto20 diversity20to33 diversityover33  aum_naturallog instcollv turnover   ftse4goodever 
> if IM==1, nolog

Random-effects logistic regression              Number of obs      =       244
Group variable: AccountName_~m                  Number of groups   =       109

Random effects u_i ~ Gaussian                   Obs per group: min =         1
                                                               avg =       2.2
                                                               max =         5

                                                Wald chi2(14)      =     21.14
Log likelihood  = -106.32473                    Prob > chi2        =    0.0980

--------------------------------------------------------------------------------------
          seekESGinf |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
---------------------+----------------------------------------------------------------
             PFbuyin |  -4.126424   1.489385    -2.77   0.006    -7.045565   -1.207283
csrcountryrating_log |   -.755804   1.769099    -0.43   0.669    -4.223174    2.711566
       indivlegitlog |  -.6593336   .5442709    -1.21   0.226    -1.726085    .4074177
         orglegitlog |  -1.004266   1.133888    -0.89   0.376    -3.226646    1.218115
        leftvotes100 |  -.0341208   .0192972    -1.77   0.077    -.0719426     .003701
       greenvotes100 |    .049475   .0949205     0.52   0.602    -.1365658    .2355157
      performorientv |  -.4067348   1.039849    -0.39   0.696    -2.444801    1.631331
     diversityupto20 |   .2004588   .5365441     0.37   0.709    -.8511483    1.252066
     diversity20to33 |  -.2559331   .5400973    -0.47   0.636    -1.314504    .8026382
     diversityover33 |  -.3642915   .6604502    -0.55   0.581     -1.65875    .9301671
      aum_naturallog |   .2916365   .1066848     2.73   0.006     .0825381     .500735
           instcollv |   2.696824   1.255802     2.15   0.032      .235497     5.15815
            turnover |  -.0857017   1.017955    -0.08   0.933    -2.080856    1.909453
       ftse4goodever |   .3342299   .8720317     0.38   0.702    -1.374921    2.043381
               _cons |  -3.153307   5.304304    -0.59   0.552    -13.54955    7.242937
---------------------+----------------------------------------------------------------
            /lnsig2u |  -12.65824   18.20029                     -48.33016    23.01368
---------------------+----------------------------------------------------------------
             sigma_u |   .0017836   .0162311                      3.20e-11    99393.44
                 rho |   9.67e-07   .0000176                      3.11e-22           1
--------------------------------------------------------------------------------------
Likelihood-ratio test of rho=0: chibar2(01) =  2.4e-05 Prob >= chibar2 = 0.498

[Carlo Lazzaro]

Sue:
I mentioned -xtologit-, not -xtlogit-.

[Sue Bowers]

Dear Carlo,
yes, of course! Sorry. Yes, I did want to try xtologit, because I read about it somewhere and it would seem ideal, however my version of stata (12) doesn't recognize the command and the search doesn't return any results. Is this a new command? Would you be able to give me any ideas on how I could get it for my version of Stata?

[Carlo Lazzaro]

Sue:
I assumed you use Stata 14(.1), otherwise, as per FAQ, you should report the version at hand.
With Stata 12, you may want to try a pooled -ologit-

[Sue Bowers]

Thank you for your suggestion. The ologit does indeed work. The results are pasted below - do they look reasonable to you?
However, running a pooled ologit I believe doesn't look at within subject variation and also doesn't take into account the fact that the data is panel data? Please correct me if I'm wrong. This model is supposed to be part of a study where the other 'main' models are cross section random effects models on panel data. Having just one pooled model that ignores the time dimension of the data doesn't really fit with my story very well. So I'm still looking for a solution.

Below I also include the results from the model that I would ideally like to run but that doesn't work (doesn't converge, is not significant, drops almost all variables) on most years in my sample. It's only significant in the final two years. Is this mostly because the number of observations is simply too small? Is there another reason? And is there any way I can overcome the problem of few observations or do I just have to give up on those years in the sample?

Again, many thanks for your help.

. ologit  seekESGinf PFbuyin csrcountryrating_log  indivlegitlog orglegitlog leftvotes100 greenvotes100 perform
> orientv diversityupto20 diversity20to33 diversityover33  aum_naturallog instcollv turnover   ftse4goodever if
>  IM==1, nolog

Ordered logistic regression                       Number of obs   =        244
                                                  LR chi2(14)     =      32.59
                                                  Prob > chi2     =     0.0033
Log likelihood = -344.04345                       Pseudo R2       =     0.0452

--------------------------------------------------------------------------------------
          seekESGinf |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
---------------------+----------------------------------------------------------------
             PFbuyin |  -.4954593   .7196458    -0.69   0.491    -1.905939    .9150205
csrcountryrating_log |  -.5951428   1.105249    -0.54   0.590    -2.761391    1.571105
       indivlegitlog |   -.150878   .2820003    -0.54   0.593    -.7035884    .4018324
         orglegitlog |  -.3229775   .3673868    -0.88   0.379    -1.043043    .3970874
        leftvotes100 |  -.0215017   .0126801    -1.70   0.090    -.0463543    .0033509
       greenvotes100 |   .0160512    .057394     0.28   0.780    -.0964389    .1285412
      performorientv |   .0615497   .7643228     0.08   0.936    -1.436496    1.559595
     diversityupto20 |   .2522929   .3576897     0.71   0.481    -.4487661    .9533518
     diversity20to33 |   .1516756    .367621     0.41   0.680    -.5688484    .8721996
     diversityover33 |  -.2051733   .4893008    -0.42   0.675    -1.164185    .7538387
      aum_naturallog |   .2452519   .0655541     3.74   0.000     .1167683    .3737355
           instcollv |  -.0095601   .6451126    -0.01   0.988    -1.273958    1.254837
            turnover |   .3318002   .7074299     0.47   0.639    -1.054737    1.718337
       ftse4goodever |   -.439235   .4974152    -0.88   0.377    -1.414151    .5356808
---------------------+----------------------------------------------------------------
               /cut1 |  -1.587669   3.731337                     -8.900955    5.725617
               /cut2 |  -.6546266    3.72866                     -7.962666    6.653413
               /cut3 |   .5712659   3.726397                     -6.732337    7.874869
               /cut4 |   3.215977   3.734382                     -4.103278    10.53523

ologit seekESGinf PFbuyin csrcountryrating_log  indivlegitlog orglegitlog leftvotes100 greenvotes100 performo
> rientv diversityupto20 diversity20to33 diversityover33  aum_naturallog instcollv turnover   ftse4goodever if 
> IM==1 & Datayear ==2007, nolog
note: leftvotes100 omitted because of collinearity
note: performorientv omitted because of collinearity
note: diversity20to33 omitted because of collinearity
note: diversityover33 omitted because of collinearity
note: aum_naturallog omitted because of collinearity
note: instcollv omitted because of collinearity
note: turnover omitted because of collinearity
note: ftse4goodever omitted because of collinearity
convergence not achieved

Ordered logistic regression                       Number of obs   =          7
                                                  LR chi2(-2)     =      14.06
                                                  Prob > chi2     =          .
Log likelihood =          0                       Pseudo R2       =     1.0000

--------------------------------------------------------------------------------------
          seekESGinf |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
---------------------+----------------------------------------------------------------
             PFbuyin |   56858.87          .        .       .            .           .
csrcountryrating_log |  -246340.6          .        .       .            .           .
       indivlegitlog |  -8457.285          .        .       .            .           .
         orglegitlog |   5000.852          .        .       .            .           .
        leftvotes100 |          0  (omitted)
       greenvotes100 |  -558.9055          .        .       .            .           .
      performorientv |          0  (omitted)
     diversityupto20 |  -3.65e-09          .        .       .            .           .
     diversity20to33 |          0  (omitted)
     diversityover33 |          0  (omitted)
      aum_naturallog |          0  (omitted)
           instcollv |          0  (omitted)
            turnover |          0  (omitted)
       ftse4goodever |          0  (omitted)
---------------------+----------------------------------------------------------------
               /cut1 |   131653.6          .                             .           .
               /cut2 |   132055.3          .                             .           .
--------------------------------------------------------------------------------------
Note: 7 observations completely determined.  Standard errors questionable.
convergence not achieved
r(430);

. ologit seekESGinf PFbuyin csrcountryrating_log  indivlegitlog orglegitlog leftvotes100 greenvotes100 performo
> rientv diversityupto20 diversity20to33 diversityover33  aum_naturallog instcollv turnover   ftse4goodever if 
> IM==1 & Datayear ==2008, nolog
note: ftse4goodever omitted because of collinearity
convergence not achieved

Ordered logistic regression                       Number of obs   =         14
                                                  LR chi2(10)     =      23.25
                                                  Prob > chi2     =     0.0098
Log likelihood =          0                       Pseudo R2       =     1.0000

--------------------------------------------------------------------------------------
          seekESGinf |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
---------------------+----------------------------------------------------------------
             PFbuyin |  -3748.548   8.95e+08    -0.00   1.000    -1.75e+09    1.75e+09
csrcountryrating_log |   7673.873   1.35e+09     0.00   1.000    -2.66e+09    2.66e+09
       indivlegitlog |   5582.253   6.29e+08     0.00   1.000    -1.23e+09    1.23e+09
         orglegitlog |  -12640.89          .        .       .            .           .
        leftvotes100 |   28.62673   1.25e+07     0.00   1.000    -2.45e+07    2.45e+07
       greenvotes100 |   224.9611   4.81e+07     0.00   1.000    -9.43e+07    9.43e+07
      performorientv |   570.4771   6.65e+08     0.00   1.000    -1.30e+09    1.30e+09
     diversityupto20 |   948.5041   6.90e+08     0.00   1.000    -1.35e+09    1.35e+09
     diversity20to33 |   451.5626   2.65e+08     0.00   1.000    -5.19e+08    5.19e+08
     diversityover33 |  -185.6939   7.61e+08    -0.00   1.000    -1.49e+09    1.49e+09
      aum_naturallog |   -.067063   1.42e+08    -0.00   1.000    -2.79e+08    2.79e+08
           instcollv |   486.2226   7.80e+08     0.00   1.000    -1.53e+09    1.53e+09
            turnover |   2376.615          .        .       .            .           .
       ftse4goodever |          0  (omitted)
---------------------+----------------------------------------------------------------
               /cut1 |  -29140.12   5.05e+08                     -9.89e+08    9.89e+08
               /cut2 |  -29063.66          .                             .           .
--------------------------------------------------------------------------------------
Note: 14 observations completely determined.  Standard errors questionable.
convergence not achieved
r(430);

. ologit seekESGinf PFbuyin csrcountryrating_log  indivlegitlog orglegitlog leftvotes100 greenvotes100 performo
> rientv diversityupto20 diversity20to33 diversityover33  aum_naturallog instcollv turnover   ftse4goodever if 
> IM==1 & Datayear ==2009, nolog

Ordered logistic regression                       Number of obs   =         39
                                                  LR chi2(14)     =      18.63
                                                  Prob > chi2     =     0.1797
Log likelihood = -34.082276                       Pseudo R2       =     0.2146

--------------------------------------------------------------------------------------
          seekESGinf |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
---------------------+----------------------------------------------------------------
             PFbuyin |    3.31145   3.322583     1.00   0.319    -3.200693    9.823593
csrcountryrating_log |   7.461646    4.63722     1.61   0.108    -1.627138    16.55043
       indivlegitlog |  -.5769087   .9131182    -0.63   0.528    -2.366588     1.21277
         orglegitlog |  -1.808917   1.835944    -0.99   0.324    -5.407302    1.789467
        leftvotes100 |   .0208028   .0379972     0.55   0.584    -.0536704    .0952759
       greenvotes100 |  -.0164889   .2015996    -0.08   0.935    -.4116168     .378639
      performorientv |   7.675969   2.973213     2.58   0.010     1.848579    13.50336
     diversityupto20 |  -.5148694   .9989433    -0.52   0.606    -2.472762    1.443024
     diversity20to33 |  -.9875413   1.154385    -0.86   0.392    -3.250093    1.275011
     diversityover33 |  -.4107049    1.45379    -0.28   0.778    -3.260081    2.438672
      aum_naturallog |   .1458315   .2059535     0.71   0.479    -.2578299    .5494929
           instcollv |  -7.103937   4.026129    -1.76   0.078    -14.99501    .7871309
            turnover |   1.418569   2.408744     0.59   0.556    -3.302483     6.13962
       ftse4goodever |  -.7968059   1.383057    -0.58   0.565    -3.507549    1.913937
---------------------+----------------------------------------------------------------
               /cut1 |  -.3511224   12.27973                     -24.41895     23.7167
               /cut2 |   .7136661   12.32753                     -23.44784    24.87518
               /cut3 |   3.983417   12.40188                     -20.32381    28.29065
--------------------------------------------------------------------------------------

. ologit seekESGinf PFbuyin csrcountryrating_log  indivlegitlog orglegitlog leftvotes100 greenvotes100 performo
> rientv diversityupto20 diversity20to33 diversityover33  aum_naturallog instcollv turnover   ftse4goodever if 
> IM==1 & Datayear ==2010, nolog

Ordered logistic regression                       Number of obs   =         75
                                                  LR chi2(14)     =      47.70
                                                  Prob > chi2     =     0.0000
Log likelihood = -65.540758                       Pseudo R2       =     0.2668

--------------------------------------------------------------------------------------
          seekESGinf |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
---------------------+----------------------------------------------------------------
             PFbuyin |     5.3865    4.86141     1.11   0.268    -4.141689    14.91469
csrcountryrating_log |   1.635832   3.893184     0.42   0.674    -5.994669    9.266333
       indivlegitlog |  -1.831866   1.161443    -1.58   0.115    -4.108252    .4445191
         orglegitlog |  -1.276043   1.904845    -0.67   0.503    -5.009471    2.457385
        leftvotes100 |  -.0645901   .0299755    -2.15   0.031    -.1233409   -.0058392
       greenvotes100 |     .08719   .1596738     0.55   0.585    -.2257649    .4001449
      performorientv |   3.883407   2.420164     1.60   0.109     -.860028    8.626842
     diversityupto20 |   .9667672   .7443002     1.30   0.194    -.4920343    2.425569
     diversity20to33 |   1.021454   .7860921     1.30   0.194    -.5192583    2.562166
     diversityover33 |   .1472461   1.062066     0.14   0.890    -1.934366    2.228858
      aum_naturallog |      .0782   .1327366     0.59   0.556     -.181959     .338359
           instcollv |  -5.089731   5.424011    -0.94   0.348     -15.7206    5.541135
            turnover |   1.812459   1.557884     1.16   0.245    -1.240937    4.865856
       ftse4goodever |     2.4959   1.508619     1.65   0.098    -.4609389    5.452739
---------------------+----------------------------------------------------------------
               /cut1 |  -10.25159   13.15193                      -36.0289    15.52571
               /cut2 |  -8.511754   13.17765                     -34.33948    17.31597
               /cut3 |  -6.334234   13.15436                     -32.11631    19.44784
--------------------------------------------------------------------------------------

. ologit seekESGinf PFbuyin csrcountryrating_log  indivlegitlog orglegitlog leftvotes100 greenvotes100 performo
> rientv diversityupto20 diversity20to33 diversityover33  aum_naturallog instcollv turnover   ftse4goodever if 
> IM==1 & Datayear ==2011, nolog

Ordered logistic regression                       Number of obs   =        109
                                                  LR chi2(14)     =      26.48
                                                  Prob > chi2     =     0.0224
Log likelihood = -156.76687                       Pseudo R2       =     0.0779

--------------------------------------------------------------------------------------
          seekESGinf |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
---------------------+----------------------------------------------------------------
             PFbuyin |  -.9578371   .9612166    -1.00   0.319    -2.841787    .9261128
csrcountryrating_log |  -.7193628   1.684007    -0.43   0.669    -4.019956    2.581231
       indivlegitlog |   .3021077   .4327411     0.70   0.485    -.5460492    1.150265
         orglegitlog |  -.5557471   .4924045    -1.13   0.259    -1.520842    .4093481
        leftvotes100 |  -.0325085   .0198423    -1.64   0.101    -.0713987    .0063817
       greenvotes100 |   .1131334   .0944406     1.20   0.231    -.0719669    .2982336
      performorientv |  -1.483323    1.23269    -1.20   0.229     -3.89935     .932704
     diversityupto20 |   .0430423   .5405413     0.08   0.937    -1.016399    1.102484
     diversity20to33 |   .0600262   .5767674     0.10   0.917    -1.070417    1.190469
     diversityover33 |  -.1106371   .7033911    -0.16   0.875    -1.489258    1.267984
      aum_naturallog |   .4345619   .1022944     4.25   0.000     .2340685    .6350553
           instcollv |   .7466973   .9229655     0.81   0.419    -1.062282    2.555676
            turnover |   .3124467   1.083001     0.29   0.773    -1.810197     2.43509
       ftse4goodever |  -1.077466    .759816    -1.42   0.156    -2.566678    .4117463
---------------------+----------------------------------------------------------------
               /cut1 |  -6.253794   6.039239                     -18.09048    5.582897
               /cut2 |   -5.29348   6.025877                     -17.10398    6.517021
               /cut3 |  -4.327777   6.014313                     -16.11561    7.460059
               /cut4 |  -2.622336   6.019608                     -14.42055    9.175878
--------------------------------------------------------------------------------------

.

[Carlo Lazzaro]

Sue:
some remarks about your last post:
- you should cluster your SEs in the -ologit- model (please, see -vce(cluster clusterid)-) to take into account serial correlation among repeated measures on the same id;
- you may add i.year to see if time plays any role in your regression. However, be aware that you already have too many predictors vs your sample size in your first model;
- eventually, I would side-track the other models, as you have too few observations, which are the cause of the problems you reported.

[Sue Bowers]

Dear Carlo,
I have incorporated your remarks and so now I have managed to get a hold of Stata13 and I ran the clustered (by subject) xtologit you suggested. The results are below and I'm quite happy with them:

Random-effects ordered logistic regression      Number of obs      =       244
Group variable: AccountName_~m                  Number of groups   =       109

Random effects u_i ~ Gaussian                   Obs per group: min =         1
                                                               avg =       2.2
                                                               max =         5

Integration method: mvaghermite                 Integration points =        12

                                                Wald chi2(15)      =     43.01
Log pseudolikelihood  = -341.64613              Prob > chi2        =    0.0002

                                (Std. Err. adjusted for 109 clusters in AccountName_num)
----------------------------------------------------------------------------------------
                       |               Robust
            seekESGinf |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-----------------------+----------------------------------------------------------------
               PFbuyin |   .0919279     .95814     0.10   0.924    -1.785992    1.969848
  csrcountryrating_log |  -1.327829   1.114841    -1.19   0.234    -3.512876    .8572189
            indivlegit |   .0228387   .0434859     0.53   0.599     -.062392    .1080694
              orglegit |  -.0479258   .0230172    -2.08   0.037    -.0930387   -.0028129
             leftvotes |  -3.009324   .8106297    -3.71   0.000    -4.598129   -1.420519
            greenvotes |   1.364536   3.899383     0.35   0.726    -6.278115    9.007186
       diversityupto20 |   .3216744   .3683281     0.87   0.382    -.4002354    1.043584
       diversity20to33 |   .1677256   .3540636     0.47   0.636    -.5262264    .8616776
       diversityover33 |   -.214817   .5124138    -0.42   0.675     -1.21913    .7894956
        aum_naturallog |   .2681329   .0602096     4.45   0.000     .1501243    .3861415
             instcollv |   1.492623   .6841186     2.18   0.029     .1517748     2.83347
         humaneorientv |    1.11081   .5272193     2.11   0.035      .077479    2.144141
        performorientp |   .4478877   .1702547     2.63   0.009     .1141946    .7815807
              turnover |   .3658671   .6359701     0.58   0.565    -.8806114    1.612346
         ftse4goodever |  -.4265194   .3939261    -1.08   0.279      -1.1986    .3455616
-----------------------+----------------------------------------------------------------
                 /cut1 |   13.21307    5.84077     2.26   0.024     1.765369    24.66077
                 /cut2 |   14.15712    5.84357     2.42   0.015     2.703929     25.6103
                 /cut3 |    15.4061   5.869435     2.62   0.009     3.902217    26.90998
                 /cut4 |   18.07048   5.927865     3.05   0.002      6.45208    29.68889
-----------------------+----------------------------------------------------------------
             /sigma2_u |   1.19e-31   2.97e-31                      8.97e-34    1.58e-29
----------------------------------------------------------------------------------------

I did also as suggested add the year variable but only one of the coefficients is significant and I'm not sure what to do with these results. When it looks like this can I draw any conclusions? Organizations were less likely to seek ESG information in 2008 because the Coef is negative and statistically significant?

. xtologit seekESGinf PFbuyin csrcountryrating_log  indivlegit orglegit leftvotes greenvotes diversityupto20 diversity20to33 diversityover33  aum_naturallog i
> nstcollv humaneorientv performorientp turnover  ftse4goodever i.Datayear if IM==1, nolog vce(cluster AccountName_num)

Random-effects ordered logistic regression      Number of obs      =       244
Group variable: AccountName_~m                  Number of groups   =       109

Random effects u_i ~ Gaussian                   Obs per group: min =         1
                                                               avg =       2.2
                                                               max =         5

Integration method: mvaghermite                 Integration points =        12

                                                Wald chi2(19)      =     74.17
Log pseudolikelihood  = -325.06034              Prob > chi2        =    0.0000

                                (Std. Err. adjusted for 109 clusters in AccountName_num)
----------------------------------------------------------------------------------------
                       |               Robust
            seekESGinf |      Coef.   Std. Err.      z    P>|z|     [95% Conf. Interval]
-----------------------+----------------------------------------------------------------
               PFbuyin |   .5384846   1.040859     0.52   0.605    -1.501562    2.578531
  csrcountryrating_log |  -1.133928   1.080017    -1.05   0.294    -3.250722    .9828658
            indivlegit |   .0583561   .0463641     1.26   0.208    -.0325159    .1492281
              orglegit |  -.0618345    .023184    -2.67   0.008    -.1072743   -.0163947
             leftvotes |  -3.688899   .8743924    -4.22   0.000    -5.402676   -1.975121
            greenvotes |   5.459208   3.891825     1.40   0.161    -2.168628    13.08704
       diversityupto20 |   .5098119   .3729107     1.37   0.172    -.2210795    1.240703
       diversity20to33 |   .3020181   .3753647     0.80   0.421    -.4336831    1.037719
       diversityover33 |   .1045167   .5558705     0.19   0.851    -.9849695    1.194003
        aum_naturallog |   .2863932    .068257     4.20   0.000      .152612    .4201745
             instcollv |    1.70692   .7554982     2.26   0.024     .2261709    3.187669
         humaneorientv |   1.328279   .5572603     2.38   0.017     .2360691    2.420489
        performorientp |   .5494559    .213388     2.57   0.010     .1312231    .9676887
              turnover |   .1732204   .6343879     0.27   0.785    -1.070157    1.416598
         ftse4goodever |  -.4148691   .3918446    -1.06   0.290     -1.18287    .3531322
                       |
              Datayear |
                 2008  |  -2.348503   1.000355    -2.35   0.019    -4.309162   -.3878437
                 2009  |   .6099866   .7931401     0.77   0.442    -.9445395    2.164513
                 2010  |   .8069777   .8029208     1.01   0.315    -.7667181    2.380674
                 2011  |  -.1310411   .8250844    -0.16   0.874    -1.748177    1.486095
-----------------------+----------------------------------------------------------------
                 /cut1 |   15.88384    6.44516     2.46   0.014     3.251557    28.51612
                 /cut2 |   16.97171   6.461631     2.63   0.009     4.307144    29.63627
                 /cut3 |   18.35709   6.500352     2.82   0.005     5.616639    31.09755
                 /cut4 |   21.09985   6.601879     3.20   0.001     8.160403    34.03929
-----------------------+----------------------------------------------------------------
             /sigma2_u |   2.06e-31   6.07e-31                      6.39e-34    6.65e-29
----------------------------------------------------------------------------------------

[Sue Bowers]

And also, yes, I am taking your last comment on board and will not be doing any cross section analysis on this data and will motivate that decision with insufficient observations. Thank you, you have helped enormously.

[Carlo Lazzaro]

Sue:
thanks for your feed-back.
I agree that -i.year- does not seem that informative; I would get a rid of in the final model.
However, and that was the core meaning of one of my statement in the previous reply, you still have too many predictors for 244 observations (please, consider that there should be 20 observations per predictor (Katz MH. Multivariable Analysis. Second Edtion. NY: Cambridge University Press, 2006: 81), even though 10 obs per predictor may sound wise enough).
Hence, I would skim the literature in your research field to see whether a more parsimonious model was developed in the past by other researchers.