如何使用次Y轴绘制带分组索引的DataFrame

Got it, let's break this down step by step—first we'll fix up the data loading code you started, then create a plot with a secondary Y-axis for your grouped WHO DataFrame. Here's how to do it:

Step 1: Complete Data Loading & Preprocessing #

Your initial code cut off before finishing the JSON normalization, so let's fill that in and clean the data to get a usable grouped DataFrame:

import pandas as pd
import urllib2
import json

# Fetch and parse the WHO data
url = "http://apps.who.int/gho/athena/data/GHO/MORT_100.json?profile=simple&filter=COUNTRY:*;CHILDCAUSE:CH6"
response2 = urllib2.urlopen(url)
response_json2 = json.loads(response2.read())

# Normalize the core data (stored in the 'fact' section of the JSON)
dfWHO2 = pd.json_normalize(response_json2['fact'])

# Map dimension codes to human-readable labels (e.g., country names instead of codes)
dimensions = pd.json_normalize(response_json2['dimension'])
dim_label_maps = {}
for _, dim in dimensions.iterrows():
    dim_code = dim['code']
    dim_label_maps[dim_code] = {cat['code']: cat['label'] for cat in dim['category']['category']}

# Add readable columns and clean numeric data
dfWHO2['COUNTRY'] = dfWHO2['dim.COUNTRY'].map(dim_label_maps['COUNTRY'])
dfWHO2['YEAR'] = dfWHO2['dim.YEAR'].map(dim_label_maps['YEAR']).astype(int)
dfWHO2['UNDER5_MEASLES_DEATHS'] = dfWHO2['Value'].astype(float)

# Set the grouped index (Country + Year) as requested
dfWHO2.set_index(['COUNTRY', 'YEAR'], inplace=True)

# Optional: Filter to a smaller set of countries for clearer plotting (adjust as needed)
selected_countries = ['India', 'Nigeria', 'Democratic Republic of the Congo']
df_filtered = dfWHO2.loc[selected_countries, ['UNDER5_MEASLES_DEATHS']].unstack(level=0)

Step 2: Plot with Secondary Y-Axis #

We'll use matplotlib to create a plot where the primary Y-axis shows absolute death counts, and the secondary Y-axis shows annual percentage changes (a common use case for dual axes). You can swap the secondary axis metric if you have another value to compare:

import matplotlib.pyplot as plt

# Initialize figure and primary axis
fig, ax1 = plt.subplots(figsize=(12, 6))

# Plot absolute death counts on primary Y-axis
color = 'tab:blue'
ax1.set_xlabel('Year')
ax1.set_ylabel('Under-5 Measles Deaths', color=color)
for country in selected_countries:
    ax1.plot(df_filtered.index, df_filtered['UNDER5_MEASLES_DEATHS'][country], marker='o', label=country)
ax1.tick_params(axis='y', labelcolor=color)
ax1.legend(loc='upper left')

# Create secondary Y-axis for percentage change
ax2 = ax1.twinx()
color = 'tab:red'
ax2.set_ylabel('Annual % Change in Deaths', color=color)
# Calculate year-over-year percentage change for each country
for country in selected_countries:
    pct_change = df_filtered['UNDER5_MEASLES_DEATHS'][country].pct_change() * 100
    ax2.plot(df_filtered.index, pct_change, marker='s', linestyle='--', label=f'{country} % Change')
ax2.tick_params(axis='y', labelcolor=color)
ax2.legend(loc='upper right')

# Finalize plot layout and title
plt.title('Measles Deaths in Under-5 Children (Absolute Counts & Yearly Change)')
fig.tight_layout()
plt.show()

Key Details: #

• Grouped Index Handling: We set ['COUNTRY', 'YEAR'] as a multi-index, then use unstack() to reshape the data so years are on the x-axis and each country is a separate series—this makes plotting grouped data straightforward.
• Secondary Y-Axis: The twinx() method creates a second axis that shares the x-axis, perfect for comparing metrics on different scales (like raw counts vs. percentage changes).
• Readability: We use distinct colors, markers, and separate legends to keep the two axes' data clear.

If you have a different secondary metric (e.g., death rates per 1000 live births), just replace the percentage change calculation with your target column from the DataFrame.

内容的提问来源于stack exchange，提问作者kiltannen