Doubling the one-sided P-value in testing independence in 2 x 2 tables against a two-sided alternative.