Criticisms:
Information on 4,790 matched pairs not
used
Pooling exchangeable matched sets not used
Frequency matching not used
G. Analysis that controls for both matched and
unmatched variables: use logistic regression on
only discordant pairs.
H. In matched follow-up studies, need only analyze
discordant pairs. In matched case-control
studies, use only discordant pairs, provided that
there are no other control variables other than
matching variables.IX. Summary(page 414)
Detailed Outline 419