Logistic Regression: A Self-learning Text, Third Edition (Statistics in the Health Sciences)

Nominal Regression

Case Processing Summary

N

NEWTYPE .00 57

1.00 45

2.00 184

Valid 286

Missing 2

Total 288

Parameter Estimates

B Std. error Wald df Sig.

NEWTYPE

.00 Intercept 1.203 .319 14.229 1 .000

AGE .282 .328 .741 1 .389

ESTROGEN .107 .307 .122 1 .727

SMOKING 1.791 1.046 2.930 1 .087

1.00 Intercept 1.882 .402 21.869 1 .000

AGE .987 .412 5.746 1 .017

ESTROGEN .644 .344 3.513 1 .061

SMOKING .889 .525 2.867 1 .090

Exp(B) 95% Confidence interval for Exp(B)

NEWTYPE Lower bound Upper bound

.00 Intercept

AGE 1.326 .697 2.522

ESTROGEN .898 .492 1.639

SMOKING .167 2.144E-02 1.297

1.00 Intercept

AGE 2.683 1.197 6.014

ESTROGEN .525 .268 1.030

SMOKING 2.434 .869 6.815

There are two parameter estimates for each independent variable and two intercepts.
The estimates are grouped by comparison. The first set compares NEWTYPE¼0to
NEWTYPE¼2. The second comparison is for NEWTYPE¼1 to NEWTYPE¼2.
With the original coding of the subtype variable, these are the comparisons of
SUBTYPE¼2 to SUBTYPE¼0 and SUBTYPE¼1 to SUBTYPE¼0 respectively.
The odds ratio for AGE¼1 vs. AGE¼0 comparing SUBTYPE¼2 vs. SUBTYPE¼0is
exp(0.282)¼1.33.

Ordinal Logistic Regression

Ordinal logistic regression is carried out in SPSS using the PLUM procedure. We
again use the cancer dataset to demonstrate this model. For this analysis, the variable

644 Appendix: Computer Programs for Logistic Regression