 Mendelian randomization methods are prone to produce false positive results when assumptions are violated. Here, the authors propose a statistical model that offers good power to detect causation between traits while controlling the false positive rate. This article was offered by Cetanyu, Yimingqin, Tianwu, and others.