Abstract:
|
The risk ratio (RR), risk difference (RD), and odds ratio (OR) are common effect measures used in epidemiology and Public Health and are often reported in the scientific literature to measure and quantify an association between an outcome (e.g., birth defect) and intervention or risk factors of interest. The RR, RD, and OR are commonly estimated using different models and as a result the inference may differ depending upon which measure of association is chosen to be presented. The RR, RD, and OR are all measures of association based on the estimated risk and as a result the inference, e.g., p-value, should be compatible regardless of which measure of association is chosen for a particular study. I define compatibility among RR, RD, and OR estimates to be such that the resulting inference is identical regardless of which measure of association is chosen to be presented. I will discuss and demonstrate that methods commonly used to estimate these measures of association lead to p-value differences for the risk ratio, risk difference, and odds ratio. In addition, I will discuss methods that may be used to achieve compatibility among these measures of association.
|