econ_survey.py

import pandas as pd
import matplotlib.pyplot as plt
import numpy as np
import cps
import seaborn as sns
sns.set_style('white')
plt.ion()

# This is a quick hack that gets the job done :/
def repulsion(arr):
    thresh = 2 # Max shift is 2 percentage points
    for i in range(len(arr)-1):
        diff = arr[i+1]-arr[i]
        if diff < thresh:
            arr[i] = arr[i+1]-thresh

    if np.diff(arr).min() < thresh:
        arr = repulsion(arr)

    return arr

def percent_format(x):
    return str(int(np.round(x))) + '%'

df = pd.read_csv('data.txt')
n = len(df)

vo = 0.0 # text vertical offset
hoa = 0.06 # horiztonal offset of agreement percent number
hoi = 0.26 # horizontal offset of issue text

# Apply repulsion, so that text doesn't overlap
df = df.sort_values('expert_agreement')
df['expert_plot_position'] = repulsion(df['expert_agreement'].copy().values)
df = df.sort_values('public_agreement')
df['public_plot_position'] = repulsion(df['public_agreement'].copy().values)

# Create columns with nicely formatted percents
df['expert_agreement_percent'] = df['expert_agreement'].apply(percent_format)
df['public_agreement_percent'] = df['public_agreement'].apply(percent_format)

plt.figure(figsize=[9,10])

for i in range(n):

    # Nicely formatted percents
    plt.text(0 - hoa, df.ix[i, 'expert_plot_position'] + vo, df.ix[i, 'expert_agreement_percent'], ha='right', va='center', fontsize=8)
    plt.text(1 + hoa, df.ix[i, 'public_plot_position'] + vo, df.ix[i, 'public_agreement_percent'], ha='left', va='center', fontsize=8)

    # Policy question text
    plt.text(0 - hoi, df.ix[i, 'expert_plot_position'] + vo, df.ix[i, 'issue'], ha='right', va='center', fontsize=8)
    plt.text(1 + hoi, df.ix[i, 'public_plot_position'] + vo, df.ix[i, 'issue'], ha='left', va='center', fontsize=8)

    # Gray lines
    plt.plot([0, 1],  [df.ix[i, 'expert_plot_position'], df.ix[i, 'public_plot_position']], linewidth=0.5, color=[0.5, 0.5, 0.5], zorder=1)

plt.scatter(np.zeros(n), df['expert_plot_position'].values, color=cps.blue, zorder=2)
plt.scatter(np.ones(n), df['public_plot_position'].values, color=cps.pink, zorder=2)

# Titles
plt.text(-0.3, 105, "Expert Agreement", ha='right', fontsize=16)
plt.text(1.3, 105, "Public Agreement", ha='left', fontsize=16)

# Clean up chart
sns.despine(bottom=True, left=True) # Remove axes
plt.gca().xaxis.set_major_locator(plt.NullLocator()) # Remove xticks
plt.gca().yaxis.set_major_locator(plt.NullLocator()) # Remove yticks
plt.gcf().set_tight_layout(True)
plt.xlim([-2.4, 3.4])
plt.ylim([-2, 108])

plt.savefig('fig_econ_poll.png', dpi=250)