Building models with SAS Enterprise Miner, SAS Factory Miner, SAS Visual Data Mining and Machine Learning or just with programming

SAS/EM Logistic regression gains table question

Reply
New Contributor
Posts: 3

SAS/EM Logistic regression gains table question

Hello, there

I have the SAS/EM logistic regression node output as below.

My questions is, Is the "% Response" predicted or actual? What is the definition of "Mean Posterior Probability"? Thanks,

Jimmy

Data Role=TRAIN Target Variable=resp_ind

                                                                                                            Mean

                                          Cum        %               Cum          Num of      Posterior

Percentile  Gain    Lift           Lift          Resp           %Resp        Obs         Probability

      5     487.332  5.87332    5.87332    5.78840     5.78840        15652       0.060346

     10     345.051  3.02761    4.45051    2.98383    4.38616       15651       0.029073

     15     269.961  2.19777    3.69961    2.16600    3.64612       15651       0.020108

     20     216.365  1.55585    3.16365    1.53335    3.11791       15652       0.015229

     25     176.433  1.16696    2.76433    1.15009    2.72436       15651       0.012035

     30     149.594  1.15399    2.49594    1.13731    2.45985       15651       0.009786

     35     126.810  0.90109    2.26810    0.88807    2.23530       15652       0.008119

     40     108.589  0.81039    2.08589    0.79867    2.05573       15651       0.006834

     45      91.824  0.57700    1.91824    0.56865    1.89050        15651       0.005847

     50      77.634  0.49920    1.77634    0.49198    1.75065        15651       0.005064

     55      64.903  0.37600    1.64903    0.37056    1.62519        15652       0.004404

     60      54.133  0.35657    1.54133    0.35142    1.51904        15651       0.003851

     65      44.570  0.29822    1.44570    0.29391    1.42480        15651       0.003373

     70      36.698  0.34358    1.36698    0.33861    1.34721        15652       0.002943

     75      29.054  0.22043    1.29054    0.21724    1.27188        15651       0.002560

     80      22.204  0.19449    1.22204    0.19168    1.20437        15651       0.002205

     85      15.778  0.12965    1.15778    0.12778    1.14104        15652       0.001874

     90      10.066  0.12966    1.10066    0.12779    1.08475        15651       0.001562

     95       4.683  0.07780    1.04683    0.07667    1.03169         15651       0.001224

    100       0.000  0.11021    1.00000    0.10862    0.98554        15651       0.000669

New Contributor
Posts: 2

Re: SAS/EM Logistic regression gains table question

Posted in reply to jcpenny2002

"% Response" is actual: the percentage of events in each percentile.

"Mean Posterior Probability" is predicted. For constructing the gains table the data ist sorted by the predicted (posterior) probability and grouped into 20 evenly sized percentiles. "Mean posterior Probability" is the mean of the predicted probability (the sorting variable) in each group.

Ask a Question
Discussion stats
  • 1 reply
  • 297 views
  • 1 like
  • 2 in conversation