bootstrap.py

from rootpy.tree import Tree, TreeModel, FloatCol
from rootpy.io import root_open
from root_numpy import root2array, array2tree
import numpy as np
from random import gauss
import random

random.seed(0)


class Sample(TreeModel):
    a = FloatCol()
    b = FloatCol()


with root_open('sample.root', 'recreate'):
    # generate toy data in a TTree
    tree = Tree('sample', model=Sample)
    for i in xrange(1000):
        tree.a = gauss(0, 1)
        tree.b = gauss(1, 1)
        tree.Fill()
    tree.write()


# read in the TTree as a NumPy array
array = root2array('sample.root')

with root_open('bootstrap.root', 'recreate') as output:
    # bootstrap 100 times
    for bootstrap_idx in xrange(100):
        print "bootstrap {0} ...".format(bootstrap_idx)
        # resample with replacement
        sample_idx = np.random.choice(len(array), size=len(array),
                                      replace=True)
        array_bootstrapped = array[sample_idx]
        # convert back to a TTree and write it out
        tree_bootstrapped = array2tree(array_bootstrapped,
                                       name='bootstrap_{0}'.format(
                                           bootstrap_idx))
        tree_bootstrapped.Write()