Python bootstrapped confidence interval

ex mean bootstrap CI.py

# *-----------------------------------------------------------------
# | PROGRAM NAME: ex mean bootstrap CI.py
# | DATE: 10/16/20
# | CREATED BY: MATT BOGARD 
# | PROJECT FILE:        
# *----------------------------------------------------------------
# | PURPOSE: example calculation of bootstrapped confidence interval for mean
# *----------------------------------------------------------------

import numpy as np
import pandas as pd

# simulate fake data
N = 100 
Location = ['Florida','Texas']
df = pd.DataFrame(columns=['State', 'Cost'])
df['State'] = np.random.choice(Location, size=N) 
df['Cost'] = np.random.uniform(50, 1000, df.shape[0])

df.Cost.describe() # check

#
# define function for creating bootstrap samples
#

def draw_bs_replicates(data,func,size):
    """creates a bootstrap sample, computes replicates and returns replicates array"""
    # Create an empty array to store replicates
    bs_replicates = np.empty(size)
    
    # pull bootstrapped samples
    for i in range(size):
        # Create a bootstrap sample
        bs_sample = np.random.choice(data,size=len(data))
        # Get bootstrap replicate and append to bs_replicates
        bs_replicates[i] = func(bs_sample)
    
    return bs_replicates


# draw bootstrap samples of the statistic or parameter estimate of interest
bs_replicates = draw_bs_replicates(df.Cost,np.mean,500)

#
# analysis
#

# original sample mean
df.Cost.mean()

# mean of bs samples
np.mean(bs_replicates)

# bootsrapped standard error
(np.std(bs_replicates))

# 95% confidence interval based on 2.5th and 97.5th percentiles
np.percentile(bs_replicates,[2.5,97.5])