parameter for the product term will drop out in the calculation of the odds ratio. The
new variable we shall create will be called CHL220 and be equal to CHL minus 220.
We shall also create a product term CATCHL220. This can be accomplished using
the dialog box: Transform!Compute Variable and then defining the new variable
or by using the following syntax:
COMPUTE chl220¼chl-220.
EXECUTE.COMPUTE cc220¼cat * chl220.
EXECUTE.Now run the same model as before, except replace CHL220 for CHL and CC220 for
the product term CC. The desired odds ratio will be justexp(b 1 ). The syntax is as
follows:
LOGISTIC REGRESSION VAR¼chd
/METHOD¼ENTER cat age chl220 ecg smk hpt ch cc220
/PRINT¼CI(95)
/CRITERIA PIN(.05) POUT(.10) ITERATE(20) CUT(.5).The output containing the parameter estimate follows:
Variables in the EquationB S.E. Wald df Sig. Exp(B)95.0% C.I. for
EXP(B)
Lower Upper
Step 1 cat 2.528 .629 16.170 1 .000 12.526 3.654 42.944
age .035 .016 4.694 1 .030 1.036 1.003 1.069
chl220 .005 .004 1.700 1 .192 .995 .986 1.003
ecg .367 .328 1.254 1 .263 1.444 .759 2.745
smk .773 .327 5.582 1 .018 2.167 1.141 4.115
hpt 1.047 .332 9.961 1 .002 2.848 1.487 5.456
ch 2.332 .743 9.858 1 .002 .097 .023 .416
cc220 .069 .014 23.202 1 .000 1.072 1.042 1.102
Constant 5.250 .960 29.906 1 .000 .005The first row of the output shows that the estimated odds ratio for CAT¼1 vs. CAT¼ 0
among those with HPT¼0 and CHOL¼220 using this new coding isexp(2.528)¼
12.526 with corresponding 95% confidence interval (3.654, 42.944).
With the NOMREG procedure, the values of the outcome are sorted in ascending
order with the last (or highest) level of the outcome variable as the reference group. If
we wish to model P(CHD¼1), as was done in the previous analysis with the
LOGISTIC REGRESSION procedure, the variable CHD must first be recoded so
that CHD¼0 is the reference group. This process can be accomplished using the
638 Appendix: Computer Programs for Logistic Regression