Poisson Regression GLM in SAS

Question

I'm attempting a Poisson Regression general linear model in SAS.

I'm an R user, so I have no idea how to do this stuff in SAS. I'll post the data, along with the code that I've attempted already:

  Game     Success     Attempts
  1        4             5 
  2        5             11
  3        5             14
  4        5             12
  5        2             7
  6        7             10
  7        6             14
  8        9             15
  9        4             12
 10        1              4
 11        13            27
 12        5             17
 13        6             12
 14        9              9
 15        7             12
 16        3             10
 17        8             12
 18        1              6
 19        18            39
 20        3             13
 21        10            17
 22        1              6 
 23        3             12

I've tried using several different codes on the data, but I keep getting errors.

This code doesn't work for the initial input:

options nocenter;

data freethrows;

input $attempt $success;

datalines;

...(this is where I put each attempt and success for each game in each row     for 23 rows)

;

run;

The example on the SAS website is the following:

data insure;

      input n c car$ age;

      ln = log(n);

      datalines;

   500   42  small  1

   1200  37  medium 1

   100    1  large  1

   400  101  small  2

   500   73  medium 2

   300   14  large  2

   ;

   run;

The GENMOD procedure is as follows:

proc genmod data=insure;

      class car age;

      model c = car age / dist   = poisson

                      link   = log

                      offset = ln;

   run;

I'd like to run a similar analysis on the freethrows.

At the moment your data does not seem to have a predictor variable. So it's rather unclear why you want to use regression. The exact correspondences are between 'Success' == 'c' and 'Attempts' == 'n', so ti would seem to be simple matter to drop those in to the SAS code. The only remaining task is to figure out what SAS expects on the RHS for NULL models. In R it would be the number '1'. — IRTFM

IRTFM IRTFM · Accepted Answer · 2015-02-15T18:36:18

Need to take out the dollar signs since those force a variable to be considered as "character" rather than numeric. Will use "Game" as the predictor variable. Try this:

data games;
  input Game Success Attempts;
  lnAtt = log(Attempts);
  datalines;
  1        4             5 
  2        5             11
  3        5             14
  4        5             12
  5        2             7
  6        7             10
  7        6             14
  8        9             15
  9        4             12
 10        1              4
 11        13            27
 12        5             17
 13        6             12
 14        9              9
 15        7             12
 16        3             10
 17        8             12
 18        1              6
 19        18            39
 20        3             13
 21        10            17
 22        1              6 
 23        3             12
;
run;

Then execute the PROC:

proc genmod data=games;
      #  remove unless you have categorical variables; class car age;
      model Success = Game  / dist   = poisson
                      link   = log
                      offset = lnAtt;
   run;

This should be a test of the effect of the "seasoning of player experience", or some such, on success, i.e a a test of a linear trend in probability of 'Success' with increasing Game count. As a check against R's results:

    summary(glm(Success ~Game, offset=log(Attempts), family="poisson", data=games) )
#---------------------
Call:
glm(formula = Success ~ Game, family = "poisson", data = games, 
    offset = log(Attempts))

Deviance Residuals: 
    Min       1Q   Median       3Q      Max  
-1.1957  -0.7962  -0.2722   0.6774   2.1110  

Coefficients:
             Estimate Std. Error z value Pr(>|z|)    
(Intercept) -0.679572   0.189074  -3.594 0.000325 ***
Game        -0.008375   0.013544  -0.618 0.536346    
---
Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1

(Dispersion parameter for poisson family taken to be 1)

    Null deviance: 18.778  on 22  degrees of freedom
Residual deviance: 18.396  on 21  degrees of freedom
AIC: 100.72

Number of Fisher Scoring iterations: 4

So the coefficient is near zero (where a positive value would have indicated an increasing probability of Success with increase in Game count), and there is really no statistical evidence of an upward trend.

Poisson Regression GLM in SAS

1 Answers