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

The computer results for this final model are

shown here. This includes the estimated regres-

sion coefficients, corresponding standard

errors, and Wald test information. The vari-

ables CATHPT and CATCHL are denoted

in the printout as CH and CC, respectively.

Also provided here is the formula for the esti-

mated adjusted odds ratio for the CAT, CHD

relationship. Using this formula, one can com-

pute point estimates of the odds ratio for differ-

ent specifications of the effect modifiers CHL and

HPT. Further details of these results, including

confidence intervals, will be provided in Chap. 7.

SUMMARY
Three stages:
(1) Variable specification
(2) Interaction
(3) Confounding/precision

Initial model: HWF model

Hierarchical backward elimination
procedure
(test for interaction, but do not test
for confounding)

Hierarchy principle

significant product term

+

retain lower order components

As a summary of this presentation, we have

recommended a modeling strategy with three

stages: (1)variable specification, (2)interac-

tion assessment, and (3)confounding assess-

mentfollowed by consideration ofprecision.

The initial model has to behierarchically well

formulated(HWF). This means that the model

must contain all lower order components of

any term in the model.

Given an initial model, the recommended

strategy involves a hierarchical backward

elimination procedurefor removing variables.

In carrying out this strategy, statistical testing

is allowed for interaction terms, but not for

confounding terms.

When assessing interaction terms, thehierar-

chy principleneeds to be applied for any prod-

uct term found significant. This principle

requires all lower order components of signif-

icant product terms to remain in all further

models considered.

EXAMPLE (continued)

Printout

Variable Coefficient S.E. Chi sq P

Intercept 4.0497 1.2550 10.41 0.0013

CAT12.6894 3.1047 16.71 0.0000

AGE 0.0350 0.0161 4.69 0.0303

CHL0.00545 0.0042 1.70 0.1923

ECG 0.3671 0.3278 1.25 0.2627

SMK 0.7732 0.3273 5.58 0.0181

HPT 1.0466 0.3316 9.96 0.0016

CH 2.3318 0.7427 9.86 0.0017

CC 0.0692 0.3316 23.20 0.0000

interaction

CH¼CATHPT and

CC¼CATCHL

RORd ¼expð 12 : 6894

þ 0 : 0692 CHL 2 : 3881 HPTÞ

Details in Chap. 7.

Vs

192 6. Modeling Strategy Guidelines