Abstract
The situation where the prevalence rates of a disease may be affected by a number of confounding factors is considered. A set of three x2 tests are proposed for directly adjusted prevalence rates, adjusted by the population composition. The three tests are a test for total equalities among the adjusted prevalence rates, a test for linearity, and a test for slope in a linear regression of the adjusted prevalence rates on the study factor intensities. The tests are given as one set and yield the estimated adjusted prevalence rates for the dose-response curve. Approximate expressions of their powers are given. It is shown that they are usually more powerful than the x2 tests performed by summing up x2 statistics in each stratum in the style of Wood (1978), and that the x2 test for slope is similar to that proposed by Mantel (1963). An application to data [Tsubota (1979, a, b)] involving the average concentrations of NO2 and the prevalence rates of persistent cough and phlegm is given.
