data a;
input gender $ residence $ IQ $ SES $ H L;
n=H+L;
cards;
Male     Rural     High      High              117  47
Male     Rural     High      Low               54  87 
Male     Rural      Low      High               29  78
Male     Rural      Low       Low               31  262
Male     Small     High      High              350  80
Male     Small     High       Low               70  85
Male     Small      Low      High               71  120
Male     Small      Low       Low               33  265
Male     Large     High      High              151   31
Male     Large     High       Low               27   23
Male     Large      Low      High               30   27
Male     Large      Low       Low               12   52
Female   Rural     High      High              102   69
Female   Rural     High       Low               52  119
Female   Rural      Low      High               32   73
Female   Rural      Low       Low               28  349
Female   Small     High      High              338   96
Female   Small     High       Low               44   99
Female   Small      Low      High               76  107
Female   Small      Low       Low               22  344
Female   Large     High      High              148   35
Female   Large     High       Low               17   39
Female   Large      Low      High               21   47
Female   Large      Low       Low                6  116
proc logistic;
class gender residence IQ SES/param=ref;
/* HIERARCHY=MULTIPLE suggests that terms can only be entered if all */
/* their lower-level interactions are included as well */
model H/n=gender|residence|IQ|SES/selection=stepwise hierarchy=multiple aggregate scale=1;
run;