RESUME LOG WITH ADDITIONAL EXAMPLES.
===================================

(I) Use of the chisq  option with RxC table 
data, with R and C > 2 (and not necessarily equal).

libname home ".";

 proc format ;
         value agedc  Low- <40 = 35
                      40-49     = 45
                      50-59     = 55
                      60-HIGH   = 65 ;
         value  salint 0 -  250 = 1
                      251 - 400  = 2
                      401 - HIGH = 3 ;
      data newceo;
          infile "ASCdata/CEO.dat";
          input age  salry  ;
             agegp = put(age, agedc.);
             salgp = put(salry, salint.); run;
    proc freq data=newceo;
       tables salgp * agegp / nocum norow nocol nopercent
                              chisq ;
       title "Formatted Table, Recoded";  run; 

                   Formatted Table, Recoded        
                   Table of salgp by agegp

    salgp     agegp

Frequency|35      |45      |55      |65      |  Total
---------+--------+--------+--------+--------+
1        |      0 |      6 |      6 |      4 |     16
---------+--------+--------+--------+--------+
2        |      5 |      7 |      8 |      2 |     22
---------+--------+--------+--------+--------+
3        |      0 |      6 |     10 |      5 |     21
---------+--------+--------+--------+--------+
Total           5       19       24       11       59

Statistics for Table of salry by agegp

 Statistic                     DF       Value      Prob
 ------------------------------------------------------
 Chi-Square                     6     10.7487    0.0965
 Likelihood Ratio Chi-Square    6     12.3880    0.0539
 Mantel-Haenszel Chi-Square     1      0.2272    0.6336
 ...
WARNING: 50% of the cells have expected counts less
than 5. Chi-Square may not be a valid test.

         Sample Size = 59


II) Use of FORMAT  to re-code the data again into 2x2 
form to illustrate "agree" option. I start by finding the 
medians for age and salry, and using them as breakpoints.

proc means data=newceo median Q1 Q3;
     var age salry; run;      

                                Lower           Upper
Variable          Median        Quartile        Quartile
--------------------------------------------------------
age           50.0000000      45.0000000      57.0000000
salry        350.0000000     250.0000000     543.0000000
--------------------------------------------------------

proc format ;
    value $age2gp  Low - 50 = "<=50" 
                  51-HIGH = ">50" ;
    value $sal2int LOW-350  = "L"
                  351 - HIGH = "H" ; 
data newceo2;
    set newceo (rename = (age = agenum salry = salnum));
    length age $ 2 salry $ 3; 
/* specifying length correctly as number of digits for character 
   variable is important if ordering is to be done 
   lexicographically !! */
    age = agenum; salry = salnum;
proc freq data=newceo2;
    tables salry * age /agree nocum nopercent norow nocol;
    format salry $sal2int. age $age2gp.;
    run;

salry     age

Frequency|<=50    |>50     |  Total
---------+--------+--------+
L        |     16 |     16 |     32
---------+--------+--------+
H        |     14 |     13 |     27
---------+--------+--------+
Total          30       29       59

Statistics for Table of salry by age

      McNemar's Test
  -----------------------
  Statistic (S)    0.1333
  DF                    1
  Pr > S           0.7150


  Simple Kappa Coefficient
  --------------------------------
  Kappa                    -0.0184
  ASE                       0.1299
  95% Lower Conf Limit     -0.2729
  95% Upper Conf Limit      0.2361

       Sample Size = 59