propensity score
This is a way to control for multiple possible confounders .
$$ p(X) = Pr(Z = 1|X) $$ Z = primary risk factor
Estimate the propensity score (relationship between Z and the X’s using logistic regression) Stratifies into quantiles on the estimated propensity score ranging from low-high Estimate the risk factor effect on the outcome within each propensity score quantiles Pool all estimates, investigate the assocation between Y and Z, controlling for all the propensity scores