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

The model statement contains the time variable (SURVTIME) followed by a back-
slash and the case status variable (MI) with the value for cases (1) in parentheses.

The output is omitted.

Polytomous Logistic Regression

Polytomous logistic regression is demonstrated with the cancer dataset using the
NOMREG procedure described previously.

The outcome variable is SUBTYPE, a three-category outcome indicating whether the
subject’s histological subtype is Adenocarcinoma (coded 0), Adenosquamous (coded
1), or Other (coded 2). The model is restated as follows:

ln

PðSUBTYPE¼gjXÞ

PðSUBTYPE¼ 0 jXÞ



¼agþ bg 1 AGEþbg 2 ESTROGENþbg 3 SMOKING;

whereg¼ 1 ; 2

By default, the highest level of the outcome variable is the reference group in the
NOMREG procedure. If we wish to make SUBTYPE¼0 (Adenocarcinoma) the
reference group, as was done in the presentation in Chap. 12, the variable SUBTYPE
must be recoded. The new variable created by the recode is called NEWTYPE and has
already been included in the SPSS datasetcancer.sav. The command syntax used for
the recoding was as follows:

RECODE

subtype

(2¼0) (1¼1) (0¼2) INTO newtype.

EXECUTE.

To run the NOMREG procedure, select Analyze!Regression!Multinomial Logis-
tic from the drop-down menus to reach the dialog box to specify the logistic model.
Select NEWTYPE from the variable list and enter it into the Dependent Variable box,
then select and enter the covariates (AGE, ESTROGEN, and SMOKING) into the
Covariate(s) box. In the main dialog box, click on OK to run the model with the
default settings.

The corresponding syntax is shown next, followed by the output generated by run-
ning the procedure.

NOMREG

newtype WITH age estrogen smoking

/CRITERIA¼CIN(95) DELTA(0) MXITER(100) MXSTEP(5) LCONVERGE(0)

PCONVERGE

(1.0E-6) SINGULAR(1.0E-8)

/MODEL

/INTERCEPT¼INCLUDE

/PRINT¼PARAMETER SUMMARY LRT.

SPSS 643