We have so far assumed that each trait is associated with at most one causal variant per locus. However, it is not unusual to observe two or more independent associations at a locus for a trait of interest [22]. In the presence of multiple independent associations, the assumption of a single variant per trait prompts the algorithm to consider only the strongest of these distinct association signals. Hence, the presence of additional associations that explain a smaller fraction of the variance of the trait, for example additional and independently associated rare variants, have a negligible impact on our computations.
