Figure 1From: A modified regression method to test publication bias in meta-analyses with binary outcomes Empirical type I error rate. Empirical type I error rate with respect to the number of included studies and different OR values and heterogeneity: a. Type I error rate of 10 tests with heterogeneity; b. Type I error rate of 10 tests without heterogeneity. Nominal significance level is 0.10.Back to article page