BMI variation with of GDP · scipython.com

A bubble chart is a type of scatter plot which can depict three dimensions of data through the position ( $$x$$ and $$y$$ coordinates) and size of the marker. The plt.scatter method can produce bubble charts by passing the marker size to its s attribute (in $(\mathrm{points})^2$ such that the area of the marker is proportional to the magnitude of the third dimension).

The files gdp.tsv, bmi_men.tsv and population_total.tsv contain the following data from 2007 for each country: the GDP per person per capita in international dollars fixed at 2005 prices, the body mass index (BMI) of men (in $\mathrm{kg/m^2}$ ) and the total population. Generate a bubble chart of BMI against GDP, in which the population is depicted by the size of the bubble markers. Beware: some data is missing for some countries.

Bonus exercise: colour the bubbles by continent using the list provided in the file continents.tsv.

Here is one approach:

bmigdp

import numpy as np
import matplotlib.pyplot as plt


def read_data(filename, data_type=float):
    """
    Read in data from filename from two-column, tab-separated format
    of country, value pairs. Cast the value to the data type specified
    by data_type and return it as values to a dictionary keyed by country.

    """

    print(f"Reading data from file {filename}")
    data = {}
    with open(filename, "r") as fi:
        for line in fi.readlines():
            fields = line.split("\t")
            if len(fields) != 2:
                # No data for this country
                continue
            country, value = fields[0], fields[1].strip()
            if not value:
                # Blank data field for this country
                continue
            data[country] = data_type(value)
    return data


# Read in the raw data
raw_pop = read_data("population_total.tsv")
raw_bmi = read_data("bmi_men.tsv")
raw_gdp = read_data("gdp.tsv")
continent = read_data("continents.tsv", str)

# Get an alphabetically-sorted list of countries which have entries in all
# of the data sets just read in.
countries = (
    set(raw_pop.keys())
    & set(raw_bmi.keys())
    & set(raw_gdp.keys())
    & set(continent.keys())
)
countries = sorted(list(countries))

# Convert the dictionaries of data to numpy arrays
pop = np.array([raw_pop[country] for country in countries])
bmi = np.array([raw_bmi[country] for country in countries])
gdp = np.array([raw_gdp[country] for country in countries])

# Set the colours of the plotted bubbles by continent
continent_colours = {
    "Africa": "b",
    "Asia": "r",
    "North America": "y",
    "South America": "m",
    "Europe": "g",
    "Oceania": "c",
}
colours = [continent_colours[continent[country]] for country in countries]

# Set the bubble sizes (circle areas) by population, scaled to 2000 pts^2
# per billion people, but with a minimum size of 16 pts^2
sizes = np.clip(pop / 1.0e9 * 2000, 16, None)

fig, ax = plt.subplots()
sct = ax.scatter(gdp, bmi, s=sizes, c=colours, edgecolor="w")
# Make the bubbles a bit transparent: some large ones overlap smaller ones
sct.set_alpha(0.75)
# The countries' GDPs span several orders of magnitude so a log-scale
# is appropriate.
ax.set_xscale("log")

ax.set_xlabel("GDP per capita")
ax.set_ylabel("Body Mass Index of men (kg/m2)")

plt.show()

The figure produced has been styled to resemble those of gapminder.org.

BMI variation with GDP on a scatter plot.

P7.4.3BMI variation with of GDP