9 Confounding and stratified analysis

Standardisation adjusts rates by reweighting. The more general approach, for any measure and any design, is stratification: divide the data into strata within which the confounder is constant (or nearly so), estimate the association within each stratum, and then decide whether the stratum estimates can be summarised by one number. This section develops the second step and the decision; the Mantel–Haenszel procedure of Section 10 gives the summary estimator in most common use.

Questions on this section

Stuck on something here? Ask below and it stays attached to this topic.