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

Parameter Estimates

Parameter B

Std.

error

95% Wald confidence

interval Hypothesis test

Lower Upper

Wald

chi-square df Sig.

(Intercept) 1.398 1.1960 3.742 .946 1.366 1 .243

birthwgt .0005 .0003 .001 .000 2.583 1 .108

gender .002 .5546 1.085 1.089 .000 1 .997

diarrhea .221 .8558 1.456 1.899 .067 1 .796

(Scale) 1

Dependent Variable: outcome
Model: (Intercept), birthwgt, gender, diarrhea

Working Correlation Matrixa

Measurement

Measurement

[month

¼1.00]

[month

¼2.00]

[month

¼3.00]

[month

¼4.00]

[month

¼5.00]

[month

¼6.00]

[month

¼7.00]

[month

¼8.00]

[month

¼9.00]

[month¼1.00] 1.000 .525 .276 .145 .076 .040 .021 .011 .006

[month¼2.00] .525 1.000 .525 .276 .145 .076 .040 .021 .011

[month¼3.00] .276 .525 1.000 .525 .276 .145 .076 .040 .021

[month¼4.00] .145 .276 .525 1.000 .525 .276 .145 .076 .040

[month¼5.00] .076 .145 .276 .525 1.000 .525 .276 .145 .076

[month¼6.00] .040 .076 .145 .276 .525 1.000 .525 .276 .145

[month¼7.00] .021 .040 .076 .145 .276 .525 1.000 .525 .276

[month¼8.00] .011 .021 .040 .076 .145 .276 .525 1.000 .525

[month¼9.00] .006 .011 .021 .040 .076 .145 .276 .525 1.000

Dependent Variable: outcome
Model: (Intercept), birthwgt, gender, diarrhea
aThe AR(1) working correlation matrix structure is computed assuming the measurements are

equally spaced for all subjects.

The output contains tables for GEE model information, GEE parameter estimates,
and the working correlation matrix. The working correlation matrix is a 9
9 matrix
with an AR1 correlation structure. The table containing the GEE parameter esti-
mates uses the empirical standard errors by default. Model-based standard errors
could also have been requested. The odds ratio estimate for DIARRHEA¼1 vs.
DIARRHEA¼0isexp(.221)¼1.247.

The SPSS section of this appendix is completed. Next, modeling with Stata software
is illustrated.

STATA

Stata is a statistical software package that has become increasingly popular in recent
years. Analyses are obtained by typing the appropriate statistical commands in the
Stata Command window or in the Stata Do-file Editor window. The commands used

648 Appendix: Computer Programs for Logistic Regression