Learning Objectives

In the Asynchronous Lecture

Discuss the components of a good quantitative visualization
An introduction to plotting using matplotlib and seaborn
An introduction to the grammar of graphics using plotnine
Delving into visualizing:
- continuous on continuous data;
- continuous on categorical data; and
- categorical on categorical data.

In the Synchronous Lecture

Cover asking questions of data
Discuss approaches to numerical and visual summaries of data
Explore trends and fits on bivariate plots.
Talk through how to detect and handle missing data
Further discuss plotting:
- exporting plots;
- customizing plots to be publishable quality.

If you have any questions while watching the pre-recorded material, be sure to write them down and to bring them up during the synchronous portion of the lecture.

Synchronous Materials

Download the country_data.csv dataset used in the lecture.
Lecture Code

Asynchronous Materials

The following tabs contain pre-recorded lecture materials for class this week. Please review these materials prior to the synchronous lecture.

Total time: Approx. 1 hour and 22 minutes

Download the gapminder.csv dataset used in the asynchronous videos.

_

Data Visualization

`matplotlib` + `seaborn`

Code from the video

import pandas as pd
import matplotlib.pyplot as plt
import seaborn as sns


# %% -----------------------------------------
# Data

# Read data in and transform initial variables
dat = (pd.read_csv("gapminder.csv")
       .eval('lngdpPercap = log(gdpPercap)')
       .eval('lnpop = log(pop)'))

dat.head()


# %% -----------------------------------------
# Matplotlib

plt.figure(figsize=(12,7))
plt.scatter(x = dat.lngdpPercap, y = dat.lifeExp, c="green", alpha=.5,s=100) 
plt.xlabel("Log GDP perCapita")
plt.ylabel("Life Expectancy")
plt.show()



# %% -----------------------------------------
# Matplotlib library is built into pandas 


dat.plot.scatter(x="lngdpPercap",y="lifeExp",figsize=(12,7),alpha=0.5,c="green",s=100)
plt.xlabel("Log GDP perCapita")
plt.ylabel("Life Expectancy")
plt.show()


# %% -----------------------------------------
# Seaborn 

plt.figure(figsize=(12,7))
sns.scatterplot(x = "lngdpPercap",y="lifeExp",
                alpha=.5,color="green",s=100,
                data = dat)
plt.xlabel("Log GDP perCapita")
plt.ylabel("Life Expectancy")
plt.show()

`plotnine`/`ggplot2`

Code from the video

import pandas as pd
from plotnine import *
import warnings
warnings.filterwarnings('ignore') # Ignore warnings

# Run into issues with plotnine here are some resolutions
#!pip install scipy==1.2 --upgrade
#!pip install scikit-misc

# Read in data 
dat = (pd.read_csv("gapminder.csv")
       .eval('lngdpPercap = log(gdpPercap)')
       .eval('lnpop = log(pop)'))


dat.head()

# %% -----------------------------------------
# Building a plotnine plot

(ggplot(dat,aes(x = "lngdpPercap",y="lifeExp",color="continent")) + 
 geom_point(size=3,alpha=.5,show_legend=False) +
 theme_bw()+
 labs(x = "Log GDP Per Capita",y="Life Expectancy",color="",title="Life Expectancy on GDP") +
 scale_color_manual(values = ["blue","steelblue","black","gold","pink"]) +
 facet_wrap("continent",scales="free",ncol=1) +
 theme(figure_size=(5,12),
       legend_position="top"))

Continuous Data

Code from the video

import pandas as pd
from plotnine import *
import seaborn as sns

import warnings
warnings.filterwarnings('ignore') # Ignore warnings

# Run into issues with plotnine here are some resolutions
#!pip install scipy==1.2 --upgrade
#!pip install scikit-misc



# Read in data 
dat = (pd.read_csv("gapminder.csv")
       .eval('lngdpPercap = log(gdpPercap)')
       .eval('lnpop = log(pop)'))


dat.head()


############ Continuous Univariate ############ 

# %% -----------------------------------------
# Histogram

# plotnine/ggplot2 
(ggplot(dat, aes(x = 'lifeExp')) +
  geom_histogram())


# Seaborn
sns.distplot(dat.lifeExp,hist=True,kde=True)



# %% -----------------------------------------
# Density Plot

# plotnine/ggplot2 
(ggplot(dat, aes(x = 'lifeExp')) +
  geom_density(fill="blue",color="black",alpha=.5)+
  xlim(0,100))


# Seaborn
sns.kdeplot(dat.lifeExp,shade=True)


# %% -----------------------------------------
############ Continuous Bivariate ############


# %% -----------------------------------------
# Scatter Plot

# plotnine/ggplot2 
(ggplot(dat, aes(x = 'lngdpPercap', y = 'lifeExp')) +
  geom_point(alpha=.5))

# Seaborn
sns.scatterplot(x = 'lngdpPercap', y = 'lifeExp',data=dat)

# %% -----------------------------------------
# Line Plot

# plotnine/ggplot2 
(ggplot(dat, aes(x = 'lngdpPercap', y = 'lifeExp')) +
  geom_line())

nig = dat.query("country == 'Nigeria'")
(ggplot(nig, aes(x = 'year', y = 'lifeExp')) +
  geom_line())

# Seaborn
sns.lineplot(x = 'lngdpPercap', y = 'lifeExp',data=dat)


# %% -----------------------------------------
# Binned Plot (density/histogram)

# plotnine/ggplot2 
(ggplot(dat, aes(x = 'lngdpPercap', y = 'lifeExp')) +
  geom_bin2d())


# Seaborn
sns.jointplot(data=dat, x="lngdpPercap", y="lifeExp", kind="hex")

Categorical Data

Code from the video

import pandas as pd
import numpy as np
from plotnine import *
import seaborn as sns

import warnings
warnings.filterwarnings('ignore') # Ignore warnings

# Run into issues with plotnine here are some resolutions
#!pip install scipy==1.2 --upgrade
#!pip install scikit-misc


# Read in data 
dat = (pd.read_csv("gapminder.csv")
       .eval('lngdpPercap = log(gdpPercap)')
       .eval('lnpop = log(pop)'))

# Create an extra categorical variable 
dat['wealthy'] = np.where(dat.lngdpPercap > 9,"yes","no")

dat.head()


############### Univariate Categorical ###############


# %% -----------------------------------------

# Bar Plot
(ggplot(dat,aes(x='continent')) +
  geom_bar())


# Ordering Bar Plot by Frequency
cont_order = dat.continent.value_counts().index.tolist()
(ggplot(dat,aes(x='continent')) +
  geom_bar() +
  scale_x_discrete(limits=cont_order))


# Adding in more categorical data 
(ggplot(dat,aes(x='continent',fill='wealthy')) +
  geom_bar() +
  scale_x_discrete(limits=cont_order))


# Dodge
(ggplot(dat,aes(x='continent',fill='wealthy')) +
  geom_bar(position="dodge") +
  scale_x_discrete(limits=cont_order))

# Seaborn
sns.catplot(x="continent", hue = "wealthy",data=dat,kind="count")


# Bivariate: category on continuous -----------------------------------------

# %% -----------------------------------------
# Box plot


# plotnine
(ggplot(dat,aes(x='continent',y = 'lifeExp')) +
  geom_boxplot())


# Flip the axis 
(ggplot(dat,aes(x='continent',y = 'lifeExp')) +
  geom_boxplot() +
  coord_flip())


# Seaborn
sns.boxplot(x='continent',y = 'lifeExp',data=dat)

sns.boxplot(x='lifeExp',y = 'continent',data=dat)


# %% -----------------------------------------
# Violin plot

# ggplot
(ggplot(dat,aes(x='continent',y = 'lifeExp')) + 
    geom_violin())


# Seaborn
sns.violinplot(x='continent',y = 'lifeExp',data=dat)

# %% -----------------------------------------
# Jitter plot

(ggplot(dat,aes(x='continent',y = 'lifeExp')) + geom_point())

# ggplot
(ggplot(dat,aes(x='continent',y = 'lifeExp',color="continent")) +
  geom_jitter(width = .25,alpha=.5,show_legend=False))


# Layer the representations
(ggplot(dat,aes(x='continent',y = 'lifeExp',color="continent")) +
  geom_jitter(width = .1,alpha=.1,show_legend=False) +
  geom_boxplot(alpha=.5,show_legend=False))


# Seaborn
sns.stripplot(x='continent',y = 'lifeExp', data=dat)

# %% -----------------------------------------

# Heatmap: category on category on continous 

# ggplot
(ggplot(dat,aes(x='continent',y = 'wealthy',fill='lifeExp')) +
  geom_tile())

# Seaborn
M = dat.pivot_table('lifeExp','wealthy','continent')
M
sns.heatmap(M)

# %% -----------------------------------------

# Fitting categories into continuous on continuous

(ggplot(dat, aes(x='lngdpPercap',y = 'lifeExp',color="continent")) +
 geom_point())

Practice

These exercises are designed to help you reinforce your grasp of the concepts covered in the asynchronous lecture material.

For the following visualizations, please use the palmerpenguins dataset.

# Data can be imported via the palmerpenguins module
# !pip install palmerpenguins

import pandas as pd
from plotnine import *
import seaborn as sns
import matplotlib.pyplot as plt
from palmerpenguins import load_penguins

dat = load_penguins()
dat.head()

_

Question 1

Plot the body mass in grams (body_mass_g) as a boxplot. Make sure the following is true:

body_mass_g is on the x-axis with a label that reads “Body Mass (Grams)”; and the y-axis is species and has no label.
Add a title that reads “Penguin Body Mass”
Each boxplot should be a different fill for each species but make sure there is no legend.

_

Answer

# plotnine solution
p =\
(
ggplot(dat,aes(x="species",y="body_mass_g",fill='species')) +
geom_boxplot(alpha=.6,show_legend=False) + # added some additional transparency
coord_flip() +
labs(x="", y = "Body Mass (Grams)",title="Penguin Body Mass")
)
p


# %% -----------------------------------------
# seaborn solution
sns.boxplot(x="body_mass_g",y="species",data=dat)
plt.xlabel("Body Mass (Grams)")
plt.ylabel("")
plt.title("Penguin Body Mass")
plt.show()

# %% -----------------------------------------

# You can also set titles using the seaborn object directly
ax = sns.boxplot(x="body_mass_g",y="species",data=dat)
ax.set_title("Penguin Body Mass")
ax.set_ylabel("")
ax.set_xlabel("Body Mass (Grams)")

Question 2

Using plotnine, generate a scatter plot where flipper length (flipper_length_mm) is on the x-axis and bill length (bill_length_mm) is on the y-axis. The points should be a different color given the species of the penguin. Also, the points should be sized differently given the body mass (body_mass_g) of the penguin.

In addition to the above:

points should be transparent (alpha=.5)
Labs should read: x-axis = “Flipper Length (MM)”, y-axis = “Bill Length (MM)”, color legend = “Penguin Species” and size legend = “Body Mass (Grams)”
The plot theme should be minimal.

_

Answer

(
    ggplot(dat,aes(x="flipper_length_mm",
                   y="bill_length_mm",
                   color="species",
                   size="body_mass_g")) +
    geom_point(alpha=.5) +
    labs(x = "Flipper Length (MM)",y="Bill Length (MM)",
         color="Penguin Species",size = "Body Mass (Grams)") +
    theme_minimal() 
)

Question 3

Using plotnine, generate barplot that contains the Penguin species on the x-axis and a count of the number of observations for each category on the y-axis. You’ll need to drop all missing observations first.

dat.dropna()

From there, do the following:

Use the fill to break up each species by sex so that two bar plots are side-by-side (hint: position = "dodge").
Alter the order so that the “Adelie” species is first, the “Gentoo” second, and the “Chinstrap” third.
Change the fill colors so that “female” is “darkorange” and “male” is “black”
x-axis label should read “Penguin Species”; the y-axis “Number of Observations”, and the legend should have no title.
Add a title that reads “Distribution of Sex by Species”.
The theme should be black and white (_bw).

_

Answer

(
    ggplot(dat.dropna(),aes(x="species",fill="sex")) +
    geom_bar(position="dodge") +
    scale_x_discrete(limits = ["Adelie","Gentoo","Chinstrap"]) +
    scale_fill_manual(values=["darkorange","black"]) +
    labs(x="Penguin Species",y="Number of Observations", fill="",
         title = "Distribution of Sex by Species") +
    theme_bw()
)

The following materials were generated for students enrolled in PPOL564. Please do not distribute without permission.

ed769@georgetown.edu | www.ericdunford.com

Back to Course Website

Interrogation Techniques
Data Visualization and Exploration

PPOL 564 | Data Science I | Foundations

Lecture Materials for Week 7

Professor Eric Dunford (ed769@georgetown.edu)
McCourt School of Public Policy, Georgetown University

Learning Objectives

Synchronous Materials

Asynchronous Materials

_

Data Visualization

Relevant Slides

`matplotlib` + `seaborn`

Code from the video

`plotnine`/`ggplot2`

Relevant Slides

Code from the video

Continuous Data

Code from the video

Categorical Data

Code from the video

Practice

_

Question 1

_

Answer

Question 2

_

Answer

Question 3

_

Answer

Back to Course Website Interrogation Techniques Data Visualization and Exploration

PPOL 564 | Data Science I | Foundations Lecture Materials for Week 7

Professor Eric Dunford (ed769@georgetown.edu) McCourt School of Public Policy, Georgetown University

Learning Objectives

Synchronous Materials

Asynchronous Materials

_

Data Visualization

Relevant Slides

matplotlib + seaborn

Code from the video

plotnine/ggplot2

Relevant Slides

Code from the video

Continuous Data

Code from the video

Categorical Data

Code from the video

Practice

_

Question 1

_

Answer

Question 2

_

Answer

Question 3

_

Answer

Back to Course Website

Interrogation Techniques
Data Visualization and Exploration

PPOL 564 | Data Science I | Foundations

Lecture Materials for Week 7

Professor Eric Dunford (ed769@georgetown.edu)
McCourt School of Public Policy, Georgetown University

`matplotlib` + `seaborn`

`plotnine`/`ggplot2`