Think outside of the box plot - code accompanying my talk in DataTLV about box plots

missleading_plots.py

nimport numpy as np
import  pandas as pd
import sys
import plotly.graph_objects as go
from plotly.subplots import make_subplots
import plotly
import plotly.express as px

import matplotlib.pyplot as plt

import random

output_folder ='datatlv'
color = "#F7931E"
method='averaged_inverted_cdf'


np.random.seed(1)
mu, sigma = 0, 4

small_sample_size = 10
medium_sample_size = 100
large_sample_size = 1000
small_sample = np.random.normal(mu, sigma, small_sample_size)
medium_sample = np.random.normal(mu, sigma, medium_sample_size)
large_sample = np.random.normal(mu, sigma, large_sample_size)
data = [small_sample, large_sample]

uniform_data = list(range(0, 101))
interval_data = [0] * 25 + [25] * 24 + [50]*3 + [75]*24 + [100] * 25

def get_basic_layout(title=''):
    return go.Layout(
        xaxis = dict(
            showgrid=False,
            zeroline=False,
            showline=True,
            mirror=True,
            linewidth=2,
            linecolor='black'),
        yaxis = dict(
            showgrid=False,
            zeroline=False,
            showline=True,
            mirror=True,
            linewidth=2,
            linecolor='black'),
        showlegend=False,
        paper_bgcolor='rgba(0,0,0,0)',
        plot_bgcolor='rgba(0,0,0,0)',
        title_x=0.5,
        font=dict(size=30),
        title=title
    )

###############################################################################################################

def get_width(sample):
  return np.log2(len(sample))/20

def build_traces(samples, names, args={}, width_func=None, output_path=None, layout=get_basic_layout(), show_text=True, only_traces=False):
    if  width_func is None:
        traces = [go.Box(y=sample, name=name, **args) for sample, name in zip(samples, names)]
    else:
        traces = []
        for sample, name in zip(samples, names):
            traces.append(go.Box(y=sample, name=name, **args, width=width_func(sample)))
    if only_traces:
        return traces
    fig = go.Figure(layout=layout, data=traces)
    if show_text:
        for sample_idx, sample in enumerate(samples):
            for x in zip(["min","q1","med","q3","max"],np.percentile(sample, [0, 25, 50, 75, 100], method=method)):
                fig.add_annotation(
                    x=0.33 + sample_idx,
                    y=x[1],
                    text=f"{x[0]}: {round(x[1], 2)}",
                    font_size=20,
                    showarrow=False
                    )

    if output_path:
        fig.write_image(output_path, scale=10)#, dpi=1200)
    return fig

basic_args =  {}
points_args = {'boxpoints':'all', 'jitter':0.3, 'pointpos':-1.8}
points_args.update(basic_args)

fig = build_traces(
    data,
    names=['sample1', 'sample2'],
    args=basic_args,
    output_path='datatlv/simple_box.png',
    layout=get_basic_layout('Simple Box Plot'),
    show_text=False)
fig.show()
fig = build_traces(
    data,
    names=['sample1', 'sample2'],
    args=basic_args,
    output_path='datatlv/simple_box.png',
    layout=get_basic_layout('Simple Box Plot'),
    show_text=True)
fig.show()
fig = build_traces(
    data,
    names=['sample1', 'sample2'],
    args=points_args,
    output_path='datatlv/simple_box_with_points.png',
    layout=get_basic_layout('Simple Box Plot with Points'),
    show_text=False)
fig.show()
fig = build_traces(
    data,
    names=['sample1', 'sample2'],
    args=basic_args,
    width_func=get_width,
    output_path='datatlv/simple_box_with_width.png',
    layout=get_basic_layout('Simple Box Plot adjusted width'),
    show_text=False)
fig.show()

#TODO - side by side with and without width?
fig = make_subplots(rows=1, cols=2)

traces = build_traces(
    data,
    names=['sample1', 'sample2'],
    args=basic_args,
    only_traces=True)
for trace in traces:
    fig.add_trace(
        trace,
        row=1, col=1
    )

traces = build_traces(
    data,
    names=['sample1', 'sample2'],
    args=basic_args,
    width_func=get_width,
    only_traces=True)

fig.add_traces(traces, row=1, col=2)
# for trace in traces:
#     fig.add_trace(
#         trace,
#         row=1, col=2
#     )
fig.update_layout(
    title="Box Plot with and without Width",
    xaxis = dict(
        showgrid=False,
        zeroline=False,
        showline=True,
        mirror=True,
        linewidth=2,
        linecolor='black'),
    yaxis = dict(
        showgrid=False,
        zeroline=False,
        showline=True,
        mirror=True,
        linewidth=2,
        linecolor='black'),
    xaxis2 = dict(
        showgrid=False,
        zeroline=False,
        showline=True,
        mirror=True,
        linewidth=2,
        linecolor='black'),
    yaxis2 = dict(
        showgrid=False,
        zeroline=False,
        showline=True,
        mirror=True,
        linewidth=2,
        linecolor='black'),
    showlegend=False,
    paper_bgcolor='rgba(0,0,0,0)',
    plot_bgcolor='rgba(0,0,0,0)',
    title_x=0.5,
    font=dict(size=30))

for d in dir(fig):
    print(d)
fig.show()
###############################################################################################################

fig = go.Figure(layout=get_basic_layout('Simple Box Plot'))
fig.add_trace(go.Box(y=uniform_data, quartilemethod="inclusive", name="sample1"))
fig.add_trace(go.Box(y=interval_data, quartilemethod="inclusive", name="sample2"))
fig.show()

fig = go.Figure(layout=get_basic_layout('Simple Box Plot with Points'))
fig.add_trace(go.Box(y=uniform_data, quartilemethod="inclusive", name="sample1"))
fig.add_trace(go.Box(y=interval_data, quartilemethod="inclusive", name="sample2"))
fig.update_traces(boxpoints='all', jitter=0.3)
fig.show()

###############################################################################################################
  
histogram = np.concatenate((
    np.random.randint(90, 100, 7), 
    np.random.randint(80, 90, 5),
    np.random.randint(70, 80, 11),
    np.random.randint(60, 70, 14),
    np.random.randint(50, 60, 17),
    np.random.randint(40, 50, 5),
    np.random.randint(30, 40, 7),
    np.random.randint(20, 30, 8),
    np.random.randint(0, 10, 26))
)


labels = ['90 to <100', '80 to <90', '70 to <80', '60 to <70', '50 to <60', '40 to <50', '30 to <40', '20 to <30', '10 to <20', '0 to  <10']
fig = go.Figure(data=[go.Histogram(y=histogram)], layout=get_basic_layout('Patients Age Histogram'))
fig.update_layout(
    yaxis = dict(
        tickmode = 'array',
        tickvals = [95, 85, 75, 65, 55, 45, 35, 25, 15, 5],
        ticktext = labels
))
fig.show()

fig = go.Figure(data=[go.Box(y=histogram, name='Patients')],layout=get_basic_layout('Patients Age Box Plot'))
fig.show()

fig = go.Figure(data=[go.Box(y=histogram, name='Patients', notched=True)],layout=get_basic_layout('Patients Age Box Plot Notched'))
fig.show()


fig = make_subplots(rows=1, cols=2)

fig.add_trace(
    go.Histogram(y=histogram),
    row=1, col=1
)

fig.add_trace(
    go.Box(y=histogram, marker_color = '#636EFA', name='Patients'),
    row=1, col=2
)
fig.update_layout(
    title="Histogram vs Box Plot",
    xaxis = dict(
        showgrid=False,
        zeroline=False,
        showline=True,
        mirror=True,
        linewidth=2,
        linecolor='black'),
    yaxis = dict(
        showgrid=False,
        zeroline=False,
        showline=True,
        mirror=True,
        linewidth=2,
        linecolor='black'),
    xaxis2 = dict(
        showgrid=False,
        zeroline=False,
        showline=True,
        mirror=True,
        linewidth=2,
        linecolor='black'),
    yaxis2 = dict(
        showgrid=False,
        zeroline=False,
        showline=True,
        mirror=True,
        linewidth=2,
        linecolor='black'),
    showlegend=False,
    paper_bgcolor='rgba(0,0,0,0)',
    plot_bgcolor='rgba(0,0,0,0)',
    title_x=0.5,
    font=dict(size=30))
fig.show()