Artifactual SUSIE posterior effect sizes in credible sets R12

[jerome-f]

Hi,

I noticed that the susie estimated posterior effect sizes are looking very artifactual. For GWAS binary traits some of the estimated posteriors are in the 100's. (example below). The FINEMAP estimates are looking correct. This is either a numerical overflow error or a file parsing error somewhere in the pipeline.

trait region v maf beta se p mean sd prob mean_99
I9_HYPERTROCARDMYOP chr11:54529068-57879829 **11:57537356:G:A** 0.00215164 **2.59323** 0.162437 2.25736e-57 **105.173308070341** 0.160440940543077 1 **105.173308070341**

Marginal estimate is 2.59 and the posterior estimate is 105.17 !!!