The problem of testing the Hardy-Weinberg law when data are sampled from several populations with different allele frequencies is considered. An odds-ratio representation of disequilibrium is proposed. This ratio is assumed to be constant across strata, an assumption that seems more reasonable than that of previous authors of a constant difference in true and expected (under equilibrium) heterozygote proportions. A score-type test which optimally combines information from the separate strata is then derived using the theory of unbiased estimating equations. Simulations show that the new test has greater power than previously proposed tests when the disequilibrium ratio is constant across strata. The method is applied to glyoxalase genotype data from populations in India.