Shuffle and relabel observations in a csv file

README.md

#Randomization script

The idea is to do bootstrap resampling on the control dataset to make sure it is appropriate as a control dataset.

#Usage

python randomize.py <input.csv> <resample_times>

This will generate resample_times files called relabel_X.

#Operation

Start by reading a csv table, and then suffle it N times, labeling the first half control, second half test - make a new csv file for each re-label

#Authors

This is the idea of Brian McDonald, for his metagenomics analyses. I (Ivan Kryukov) have implemented a basic framework for the randomization, and assume no responsibility in how it is used.

randomize.py

#!/usr/bin/python2
import sys
import random
import csv

#Generate two columns of random variables, called 'a' and 'b'
def generate_data(variables = ['a','b'],
        samples = 10,
        generator = random.gauss,
        generator_args = {'mu':1, 'sigma':2}):
    data = [[{v:generator(**generator_args)} for v in variables] for s in range(samples)]
    return data

#Read data from a csv file
def read_data(input_file):
    data = []
    with open(input_file) as csvfile:
        reader = csv.DictReader(csvfile)
        for row in reader:
            data.append(row)
    return data

#Write data to a csv file
def write_data(data,output_file, labels = ['a','b','label']):
    with open(output_file,'w') as csvfile:
        writer = csv.DictWriter(csvfile, labels)
        writer.writeheader()
        for d in data:
            writer.writerow(d)

#Shuffle data; label first half labels[0], second - labels[1]
def relabel_half(data, labels = ['test','control']):
    random.shuffle(data)
    half = len(data)/2
    for i,d in enumerate(data):
        if i < half:
            d['label'] = labels[0]
        else:
            d['label'] = labels[1]
    return data

if __name__ == '__main__':
    input_file = sys.argv[1]
    data = read_data(input_file)
    for i in range(int(sys.argv[2])):
        r = relabel_half(data)
        write_data(r, "relabel_"+str(i))