Which Test?

Draft: This page has not been reviewed by a statistician yet. Use it as a starting point and check important choices with a statistician.

Conditional logistic regression

Estimates an odds ratio in matched case-control data, adjusted for other variables, by comparing within matched sets.

Clinical example

A case-control study in which every case is matched to two controls by age and sex, asking whether smoking is associated with the disease.

When to use it

  • The outcome is yes or no, for example case or control.
  • Cases and controls are matched, for example by age and sex.
  • You want an odds ratio, adjusted for other variables.

Assumptions to check

Matched sets are identified

Each case and its matched controls share a set identifier. The analysis compares within sets.

Independent matched sets

Different sets contain different patients.

What to report

  • Adjusted OR with 95% CI

Common mistakes

Ordinary logistic regression, which ignores the matching

Alternatives

If your situation is…Use
Case-control (not matched)Logistic regression

In your software

R
# R
library(survival)
fit <- clogit(case ~ smoking + bmi + strata(match_id), data = d)
summary(fit)
SPSS
* SPSS has no conditional logistic command. A known workaround is a Cox
* regression with the matched set as strata and a constant time variable.
COMPUTE time = 1.
EXECUTE.
COXREG time
  /STATUS=case(1)
  /STRATA=match_id
  /METHOD=ENTER smoking bmi.

Sources