Summary

Multivariate Data Split - Program to create subsets of a data sets that are likely to be not statistically significantly different.

Description

Subsets are picked based on the psuedo gower distance, where the distance between subjects picked in each iteration is minimized.

Input

a) phenotype file (only numerical values, i.e. please convert information such as sex (M/F) into 0/1)

b) phenotype list to be used for comparison (see example phenotype_list.csv file)

c) Number of subsets

Output

CSV file with added column including subset IDs.

Execution

./subset_data.py INFILE OUTFILE NUM_GROUPS (PHENOTYPE_LIST)

Example

./subset_data.py phenotypes.csv phenotypes_split.csv 5 phenotype_list.csv

Requires

numpy, scipy

Citations

Schirmer et al. "Rich-Club organization: an important determinant of functional outcome after acute ischemic stroke." BioRxiv (2019): 545897.