Skip to Main Content

chapter9.htm


SAS Textbook Examples

An Introduction to Categorical Analysis by Alan
Agresti
Chapter 9 – Models for Matched Pairs

9.1
Comparing Dependent Proportions

9.1.1
McNemar Test

In SAS proc freq we can request McNemar test by using
option agree after the tables statement. Notice that in the book
the McNemar test is done in terms of z-statistic, while in SAS it is done as a
chi-square statistic. 

data rating;
  input rating1 rating2 count;
cards;
1 1 794
1 0 150
0 1 86
0 0 570
;
run;
proc freq data = rating;
  weight count;
  tables rating1*rating2 /agree;
run;
The FREQ Procedure

Table of rating1 by rating2

rating1     rating2

Frequency|
Percent  |
Row Pct  |
Col Pct  |       0|       1|  Total
---------+--------+--------+
       0 |    570 |     86 |    656
         |  35.63 |   5.38 |  41.00
         |  86.89 |  13.11 |
         |  79.17 |   9.77 |
---------+--------+--------+
       1 |    150 |    794 |    944
         |   9.38 |  49.63 |  59.00
         |  15.89 |  84.11 |
         |  20.83 |  90.23 |
---------+--------+--------+
Total         720      880     1600
            45.00    55.00   100.00

Statistics for Table of rating1 by rating2

     McNemar's Test
------------------------
Statistic (S)    17.3559
DF                     1
Pr > S            <.0001

    Simple Kappa Coefficient
--------------------------------
Kappa                     0.6996
ASE                       0.0180
95% Lower Conf Limit      0.6644
95% Upper Conf Limit      0.7348

Sample Size = 1600

9.1.2
Estimating Differences of Proportions

9.2
Logistic Regression for Matched Pairs

9.2.1
Connection between McNemar and Cochran-Mantel-Haenszel Tests