Visualizing \(h^2\) results: Univariate LDSR

Comparison of our primary \(h^2\) results from partitioned LDSR to results from univariate LDSR with default settings. These differ in two key ways:

• Reference panel: Univariate 1000 Genomes European LD scores (i.e. precomputed ./eur_w_ld) vs. BaselineLD v1.1 LD scores from 1000 Genomes Europeans, as described in Gazal et al. 2017 (also on biorxiv)
• Max chi^2 filtering: Default settings with “two-step” estimation (limiting intercept estimation to regression with \(\chi^2 < 30\) before estimating \(h^2\) in regression with the intercept fixed to the first stage estimate) vs. one-step estimation with no \(\chi^2\) filter