Each simulated dataset was analyzed using mash, following the procedure described above. Note that we re-estimated (U,π) from the data without making use of the true values used to simulate the data. We estimated effects by their posterior means (eq. 8), and assessed significance by the lfsr. The ash and mash-bmalite methods were applied to the simulated data in the same way that they were applied to the GTEx data (see above).
