ECON 0150 | Economic Data Analysis

The economist’s data analysis skillset.

Part 1.1 | Summarizing Categorical Variables

We cannot typically understand our data without summarizing it.

The main differentiator between a good and a bad summarization tool is whether it’s appropriate for the data.

Summarizing Categorical Variables

… use the appropriate summary tool for the variable type

Catagorical Variables: Visualizations

Q. Which state has the most locations?

STATE
CA
NY
TX
WA
FL
...

> it’s difficult to see from the data alone

Catagorical Variables: Visualizations

Q. Which state has the most locations?

> pay attention to which of these two figures is easier to answer the question

> it’s pretty easy to see that it’s CA from both of these figures

Catagorical Variables: Visualizations

Q. Does FL or WA have more shops?

> pay attention to which of these two figures is easier to answer the question

> a bar graph is much easier to read

Catagorical Variables: Visualizations

Q. How many shops are in FL?

> pay attention to which of these two figures is easier to answer the question

> now it takes a second to read the bar graph…

Catagorical Variables: Bar Plots

Q. How many shops are in FL?

> pay attention to which of these two figures is easier to answer the question

> we can make the bar graph easier to read by placing the number near the bar

Catagorical Variables: Remove Clutter

Q. How many shops are in the state with the second most locations?

> removing clutter guides your eye to the important information

Catagorical Variables: Remove Clutter

Q. How many shops are in the state with the second most locations?

> removing clutter guides your eye to the important information

Catagorical Variables: Order by Size

Q. How many shops are in the state with the second most locations?

> states have no inherent order, but sorting can make comparisons easier

Binary Categorical Variables: CA vs Other

Q. How does CA compare to the whole?

> instead of a nominal categorical variable, this is binary (CA / Other)

Binary Categorical Variables: Binary Visualization

Q. How does CA compare to the whole?

> this question is much easier to see when visualizing the two categories

> here both the pie and the bar communicte the data effectively

Binary Categorical Variables: Percentages

Q. How does CA compare to the whole?

> if the question is about percentages, a pie chart may work best

Takeaways

… use the right summary tool for the variable type

• Binary Categorical Variables: use a pie chart or bar graph
• Nominal Categorical Variables: use a bar graph; maybe order by value
• Ordinal Categorical Variables: use an ordered bar graph

• Remove clutter; keep it simple
• Place information near the object it describes

The Framework: Select, Transform, Encode

Every visualization follows three steps

• SELECT — Which rows are we looking at?
• TRANSFORM — How do we summarize or reshape the data?
• ENCODE — How do we map values to visual elements?

S-T-E for Categorical Variables

What we just did

Step	Action
SELECT	All coffee shops
TRANSFORM	Count by state
ENCODE	Category → position; Count → bar length

> for categorical variables, TRANSFORM almost always means counting

Exercise 1.1 | Categorical Variables

Lets visualize coffee shops by state.

• Dataset 1: Coffee_Shops.csv

Lets visualize the main variable in each dataset.

• Dataset 2: employment_status.csv
• Dataset 3: household_savings.csv
• Dataset 4: household_incomes.csv

Exercise 1.1: Dataset 1

Summarize Coffee_Shops.csv as a nominal categorical variable.

# Load Dataset
shops = pd.read_csv(file_path + 'Coffee_Shops.csv')

# Summary Table
shops.value_counts()

# Countplot (bar plot)
sns.countplot(data=shops, y='STATE', hue='STATE')

Exercise 1.1: Dataset 1

Summarize Coffee_Shops.csv as a binary categorical variable.

# Load Dataset
shops = pd.read_csv(file_path + 'Coffee_Shops.csv')

# Create a binary categorical variable
shops['CA'] = np.where(shops['STATE'] == 'CA', 'CA', 'Other')

# Countplot
sns.countplot(data=shops, y='CA', hue='CA')

Exercise 1.1: Dataset 2

Summarize employment_status.csv.

Exercise 1.1: Dataset 3

Summarize household_savings.csv.

Exercise 1.1: Dataset 4

Summarize household_incomes.csv.