Contents:

      Intro

      Code

      Reproduce the simulations

      Reproduce the gene dosage experiment





Intro


This website accompanies the paper:
  • Accumulation tests for FDR control in ordered hypothesis testing.
    Ang Li and Rina Foygel Barber. arXiv:1505.07352 (To appear in JASA.)

This paper proposes a family of accumulation tests for ordered hypothesis testing, where we seek to test a ranked list of n hypotheses, H1, ..., Hn, accompanied by p-values p1, ..., pn. The main idea is to estimate the false discovery proportion among the top k hypotheses in the ranked list, for each k=1,...,n, to determine an adaptive cutoff for the list. We aim to find a cutoff k that enables us to discover as many true signals as possible while controlling the false discovery rate. For more information, see the paper.

Here we give code in the R language providing an implementation of the accumulation test method, and reproducing the empirical results in the paper.





Code


Implementation:
Demo: we create a sequence of hypotheses and p-values, and running accumulation test methods to test these hypotheses. This demo is implemented in R Markdown.





Reproduce the simulations


Here we reproduce the simulated data results in the paper, implemented in R Markdown.

Simulated sequences:

Simulated regression data:







Reproduce the gene dosage data experiment


Here we reproduce the gene dosage data results in the paper, implemented in R Markdown.