pip install dabest

import pandas as pd
import matplotlib.pyplot as plt
import dabest

# Load the iris dataset. Requires internet access.
iris = pd.read_csv("https://github.com/mwaskom/seaborn-data/raw/master/iris.csv")

# Load the above data into `dabest`.
iris_dabest = dabest.load(data=iris, x="species", y="petal_width",
                          idx=("setosa", "versicolor", "virginica"))

# Produce a Cumming estimation plot and save plot.
iris_dabest.mean_diff.plot(fig_size=(5,7))
plt.savefig("quick_start.png")

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
import dabest

np.random.seed(20190623)

# Generate toy dataset
x1 = np.random.normal(loc=50, scale=10, size=50)
x2 = np.random.normal(loc=60, scale=5, size=40)
df = pd.DataFrame(data=[x1, x2]).T
df.columns = ['Control', 'Test']

# Load the above data into `dabest`.
df_dabest = dabest.load(data=df, idx=("Control", "Test"))
df_dabest.mean_diff.plot(fig_size=(5,4))
#df_dabest.cohens_d.plot(fig_size=(5,4))
plt.savefig("ga1.png", bbox_inches="tight")

# Generate toy dataset
size = 20
idx = np.arange(0, size)
x1 = np.random.normal(loc=50, scale=10, size=size)
x2 = np.random.normal(loc=60, scale=10, size=size)
df = pd.DataFrame(data=[idx, x1, x2]).T
df.columns = ['idx', 'Control', 'Test']

# Load the above data into `dabest`.
df_dabest = dabest.load(data=df, idx=("Control", "Test"),
                        paired=True, id_col="idx")

df_dabest.cohens_d.plot(fig_size=(4,4))
#df_dabest.mean_diff.plot(fig_size=(4,4))
plt.savefig("ga2.png", bbox_inches="tight")

# Generate toy dataset
size1 = 20
size2 = 25

xc1 = np.random.normal(loc=50, scale=20, size=size1)
xt1 = np.random.normal(loc=55, scale=15, size=size1)
xc2 = np.random.normal(loc=60, scale=15, size=size2)
xt2 = np.random.normal(loc=50, scale=25, size=size2)
df = pd.DataFrame(data=[xc1, xt1, xc2, xt2]).T
df.columns = ['Control1', 'Test1', 'Control2', 'Test2']

# Load the above data into `dabest`.
df_dabest = dabest.load(data=df, idx=(("Control1", "Test1"),
                                      ("Control2", "Test2")))

#df_dabest.cohens_d.plot(fig_size=(5,5))
#df_dabest.mean_diff.plot(fig_size=(5,5))
df_dabest.hedges_g.plot(fig_size=(5,5))
plt.savefig("cumming.png", bbox_inches="tight")

import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns

np.random.seed(20190623)

# generate observation (toy data)
size = 50 # sample size
sigma = 20
mu1 = 50
mu2 = 60
diff_mu = abs(mu2 - mu1)

x1_obs = np.random.normal(loc=mu1, scale=sigma, size=size)
x2_obs = np.random.normal(loc=mu2, scale=sigma, size=size)

num_bootstrap = 5000
alpha = 0.025

es_bootstrap = np.zeros(num_bootstrap) # array to save bootstrapped effect size

for i in range(num_bootstrap):
    # bootstrapping
    x1_resampling = np.random.choice(x1_obs, size=size, replace=True)
    x2_resampling = np.random.choice(x2_obs, size=size, replace=True)

    # calculate effect size (mean difference) and save.
    mu = np.mean(x2_resampling) - np.mean(x1_resampling)
    es_bootstrap[i] += mu

# plot
ax = sns.kdeplot(es_bootstrap, shade=True)

data_x, data_y = ax.lines[0].get_data()
plt.vlines(0, ymin=0, ymax=data_y.max(), color='k', linestyle='dashed')
plt.vlines(diff_mu, ymin=0, ymax=data_y.max(), color='r', linestyle='dashed')
plt.vlines(np.mean(es_bootstrap), ymin=0, ymax=data_y.max(), color='g')
es_bootstrap_sorted = np.sort(es_bootstrap) # to plot 95% CI
idx_alpha = np.argmin(np.abs(data_x-es_bootstrap_sorted[int(num_bootstrap*alpha)]))
idx_one_m_alpha = np.argmin(np.abs(data_x-es_bootstrap_sorted[int(num_bootstrap*(1-alpha))]))
plt.vlines(es_bootstrap_sorted[int(num_bootstrap*alpha)], ymin=0, ymax=data_y[idx_alpha])
plt.vlines(es_bootstrap_sorted[int(num_bootstrap*(1-alpha))], ymin=0, ymax=data_y[idx_one_m_alpha])
plt.xlabel("Mean difference")
plt.ylabel("Density")
#plt.show()
plt.savefig("bootstrap_ci.png")