Multivariate Bayesian variable selection regression

Calibration of SNP level PIP

Here for susie and DAP we group all PIP across all simulations into 10 bins according to reported PIP. Then compute the proportion of true signals in each bin. We align this proportion to the average PIP for each bin for calibrated SNP-level posterior probability.

Calibrated PIP: for example for SNPs having PIP = 0.5, indeed half of them are true signal (simulated).

In [2]:
%revisions -s
Revision Author Date Message
0a2b456 Gao Wang 2018-06-05 Polish calibrated PIP plot
95d3072 Gao Wang 2018-06-05 Update documentation
00da2df Gao Wang 2018-06-05 Add more PIP comparisons

Workflow

Run the cali_pip workflow in this notebook

In [3]:
%cd ~/GIT/github/mvarbvs/dsc

/home/gaow/GIT/github/mvarbvs/dsc

susie with var(Y) as residual

A bit conservative.

In [7]:
%preview susie_comparison/PIP_comparison_0622.calibrated.estvar_false.png
> susie_comparison/PIP_comparison_0622.calibrated.estvar_false.png (327.9 KiB):

susie with estimated residual

A bit anti-conservative?

In [8]:
%preview susie_comparison/PIP_comparison_0622.calibrated.estvar_true.png
> susie_comparison/PIP_comparison_0622.calibrated.estvar_true.png (328.8 KiB):
Copyright © 2016-2020 Gao Wang et al at Stephens Lab, University of Chicago