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

Chapter 6: Modeling Strategy
Guidelines

All confidence intervals are quite wide, indicat-

ing that their corresponding point estimates

have large variances. Moreover, if we focus on

the three confidence intervals corresponding

to HPT equal to 1, we find that the interval

corresponding to the estimated odds ratio of

0.31 lies completely below the null value of 1.

In contrast, the interval corresponding to the

estimated odds ratio of 1.22 surrounds the null

value of 1, and the interval corresponding to

4.89 lies completely above 1.

From a hypothesis testing standpoint, these

results therefore indicate that the estimate of

1.22 is not statistically significant at the 5%

level, whereas the other two estimates are sta-

tistically significant at the 5% level.

We suggest that the reader review the material

covered here by reading the summary outline

that follows. Then you may work the practice

exercises and test.

In the next chapter, “Modeling Strategy Guide-

lines”, we provide guidelines for determining a

best model for an exposure–disease relation-

ship that adjusts for the potential confounding

and effect-modifying effects of covariables.

EXAMPLE

Wide CIs)estimates have large

variances

HPT¼1:

dOR¼0.31, CI: (0.10, .91) below 1

dOR¼1.22, CI: (0.48, 3.10) includes 1

dOR¼4.89, CI: (1.62, 14.52) above 1

ORd

(Two-tailed)

significant?

CHL¼200: 0.31 Yes

220: 1.22 No

240: 4.89 Yes

SUMMARY

Chapter 5: Statistical Inferences
Using ML Techniques

This presentation is now complete. In sum-

mary, we have described two test procedures,

the likelihood ratio test and the Wald test. We

have also shown how to obtain interval esti-

mates for odds ratios obtained from a logistic

regression. In particular, we have described

confidence interval formula for models with

and without interaction terms.

Presentation: VIII. Numerical Example 153