Note to self

sample.py

import pandas as pnd
import numpy as np
import numpy.random as rnd

from collections import namedtuple

class Sampler(list):

	def __init__(self, items, weights=None):
		super(Sampler, self).__init__(items)
		_weights = np.array(weights) if weights else np.ones(len(self))
		self.probabilities = _weights / _weights.sum()

	def sample(self, size=1):
		indices = rnd.choice(len(self), size, p=self.probabilities)
		output = [self[index] for index in indices]
		return output[0] if size == 1 else output

def simulate(mixture, num_sims, sim_horizon):
	columns = pnd.MultiIndex.from_product([['samples', 'names'], range(num_sims)])
	df = pnd.DataFrame(columns=columns)
	for sim in xrange(num_sims):
		for t in xrange(sim_horizon):
			process = mixture.sample()
			df.loc[t, [('names', sim), ('samples', sim)]] = [process.name, process.sample()]
	return df

process = namedtuple("process", ["name", "sample"])

mixture = Sampler([
	process("good", lambda : rnd.bernoulli(0.9)), 
	process("bad", lambda : rnd.bernoulli(0.5))
])

df = simulate(mixture, 10, 100)