Activity Number:
|
50
|
Type:
|
Contributed
|
Date/Time:
|
Sunday, August 11, 2002 : 4:00 PM to 5:50 PM
|
Sponsor:
|
Section on Statistical Education*
|
Abstract - #301857 |
Title:
|
Associations Involving Binary Variables: Spuriousity, Relative Risk, Phi and Sufficiency
|
Author(s):
|
Milo Schield*+ and Thomas Burnham
|
Affiliation(s):
|
Augsburg College and Cognitive Consulting
|
Address:
|
2211 Riverside Drive, Minneapolis, Minnesota, 55454, US
|
Keywords:
|
Epidemiology ; critical thinking ; necessity ; evidence ; confounding ; correlation
|
Abstract:
|
(1) The relationships between three binary variables are examined when the association between two of them is completely spurious. Cornfield's inequality and Gastwirth's equation are confirmed. Additional equations are generated and their limits of applicability are derived. (2) The relationship between relative risk and correlation is examined for two binary variables. An evaluation is made of the epidemiological use of relative risk when Phi< < 1. If the Phi for a given prevalence is expressed relative to the maximum correlation for that prevalence, this relative correlation is found to equal the Attributable Fraction in the Population, in which relative risk is a dominate factor. (3) Quantitative measures for sufficiency (S) and necessity (N) are introduced and related formulas are derived. Using these relationships, a Bayesian comparison and an over-involvement comparison are introduced. Although relative risk is proportional to sufficiency, it is found to measure necessity for a given prevalence. (4) Common measures of association between two binary variables are defined. Formulas relating these measures are derived along with formulas involving non-response bias.
|