Using simulation to compare sensitivity of PLIS procedure and conventional p-value procedure
This is a detailed demo of a simulation to test the sensitivity of PLIS procedure and conventional p-value procedure using logistics regression. While PLIS procedure incorporates the correlation among SNPs to adjust for Linkage Disequilibrium (LD), conducting multiple logistics regressions assumes that SNPs are independent from one another.
The main idea here:
We don’t have info for all SNPs along our chromosomes for research. Thus, when missing the data for true causal SNPs, we want to detect the SNPs that are physically close to the causal SNPs so that we can dig into those areas later.
When using PLIS procedure, we are incorporating SNPs’ correlations with one another, so we expect the SNPs that are physically close to the causal SNPs can detected more efficiently than when using conventional p-value procedure.
We want to see how sensitive PLIS vs. p-values are toward SNPs that are physically close to the causal SNPs