adafdr 0.0.6

A fast and covariate-adaptive method for multiple hypothesis testing

AdaFDR

A fast and covariate-adaptive method for multiple hypothesis testing.

Software accompanying the paper "AdaFDR: a Fast, Powerful and Covariate-Adaptive Approach to Multiple Hypothesis Testing", 2018.

Installation

pip install adafdr

Usage

adafdr mainly offers two methods: adafdr_explore for covariate visualization and adafdr_test for multiple hypothesis testing.

Import package and load data

adafdr.method contains the algorithm implementation while adafdr.data_loader can be used to load the data used in the paper. Here we load the airway data used in the paper. See vignette for other data accompanied with the package.

import adafdr.method as md
import adafdr.data_loader as dl
p,x = dl.data_airway()

The data p,x has the following format:

• p: (N,) numpy.ndarray, p-values for N hypotheses.
• x: (N,d) numpy.ndarray, d-dimensional covariate for each hypothesis. When d=1, x is allowed to be (N,) numpy.ndarray or (N,1) numpy.ndarray.

Covariate visualization using adafdr_explore

md.adafdr_explore(p, x, output_folder=None)

If output_folder is a folder path, figures will be saved to the folder instead of being plotted in the console.

Here, the left is a scatter plot of each hypothesis with p-values (y-axis) against the covariate (x-axis). The right are the estimated null hypothesis distribution (blue) and the estimated alternative hypothesis distribution (orange) with respect to the covariate. Here we can conclude that a hypothesis is more likely to be significant if the covariate (gene expression) value is larger.

Multiple hypothesis testing using adafdr_test

n_rej,t_rej,theta = md.adafdr_test(p, x, fast_mode=True, output_folder=None)

• If fast_mode is True, AdaFDR-fast is used, otherwise, AdaFDR is used.
• If output_folder is a folder path, log files will be saved in the folder.
• n_rej is the number of rejections, t_rej is a (N,) numpy.ndarray for decision threshold for each hypothesis, theta is a list of learned parameters.

Here, the learned threshold looks as follows. Note that the two lines correspond to the data from two folds via hypothesis splitting

Quick Test

Here is a quick test. First check if the package can be succesfully imported:

import adafdr

Next, run a small example which should take a few seconds:

import numpy as np
p,x,h,_,_ = adafdr.data_loader.load_1d_bump_slope()
n_rej,t_rej,theta = adafdr.method.adafdr_test(p, x, fast_mode=True)
D = np.sum(p<=t_rej)
FD = np.sum((p<=t_rej)&(~h))
print('# AdaFDR successfully finished! ')
print('# D=%d, FD=%d, FDP=%0.3f'%(D, FD, FD/D))

It runs AdaFDR-fast on a 1d simulated data. If the package is successfully imported, the result should look like:

# AdaFDR successfully finished! 
# D=840, FD=80, FDP=0.095

Citation information

Coming soon.