I'm interested in using COJO and trying to understand more about how the sample size difference is handled in COJO. Specifically, I have questions about equation 14 of the COJO paper (https://www.nature.com/articles/ng.2213).
Q1: There is a "2" after "min(...)" term. I was wondering if it means to square the min Neff or means to represent the squared root or a typo (?).
Q2: My intuition about B_jk in eqn 14 is that B_jk is approximately min(n_j, n_k) * cov(X_j, X_k). If so, I don't know how to think about taking inverse of B (a bit counter-intuitive to me) since the Neff is different for different entries in B. Could you point me to any references that justifies this approach (if any)?