Detailed Outline
Abbreviated Outline
I. Overview(pages 206–207)
Focus:
Assessing confounding and interaction
Obtaining a valid estimate of theE–Drelationship
A. Three stages: variable specification, interaction
assessment, and confounding assessment
followed by consideration of precision.
B. Variable specification stage
i. Start withD, E, andC 1 ,C 2 ,...,Cp.
ii. ChooseVs fromCs based on prior research or
theory and considering potential statistical
problems, e.g., collinearity; simplest choice is
to letVsbeCs themselves.
iii. ChooseWs fromCs to be eitherVs or product
of twoVs; usually recommendWstobeCs
themselves or some subset ofCs.
C. The model must behierarchically well formulated
(HWF): given any variable in the model, all lower
order components must also be in the model.
D. The strategy is ahierarchical backward
elimination strategy: evaluateEViVjterms first,
thenViterms, thenViterms last.
E. Thehierarchy principleneeds to be applied for any
variable kept in the model: If a variable is to be
retained in the model, then all lower order
components of that variable are to be retained in
all further models considered.
II. Interaction assessment stage(pages 207–210)
A. Flow diagram representation.
B. Description of flow diagram: test higher order
interactions first, then apply hierarchy principle,
then test lower order interactions.
C. How to carry out tests: chunk tests first, followed
by backward elimination whether or not chunk
test is significant; testing procedure involves
likelihood ratio statistic.
D. Example.
III. Confounding and precision assessment when no
interaction(pages 211–215)
A. Monitor changes in the effect measure (the odds
ratio) corresponding to dropping subsets of
potential confounders from the model.
B. Gold standard odds ratio obtained from model
containing allVs specified initially.
C. Identify subsets ofVs giving approximately the
same odds ratio as gold standard.
Detailed Outline 233
