8.4 Choosing between the methods
|
| Direct | Indirect |
| Needs | stratum rates of each study population; a standard population | total deaths in each study population; standard rates |
| Weights | the standard’s structure | each study population’s own structure |
| Comparable between study populations | yes | only if rate ratios are constant across strata |
| Stability with small numbers | poor | good |
Both summaries are adequate when the stratum ratios are roughly constant, and both hide something important when they are not. In that case the proper report is the stratum-specific comparison, with a summary only as a supplement. The Mantel–Haenszel methods of the next sections give pooled estimates with better statistical properties under the assumption of a common ratio, and Section 13 sets them side by side with standardisation on one set of data.
Questions on this section
Stuck on something here? Ask below and it stays attached to this topic.