Data Visualization and Exploration

datawrangling, at last

Plan for final

Go find some data!

• kaggle
• github
• data.gov
• OSF
• dryad.org
• World Bank
• NSW Health and additional resources
• Or your OWN data!

Check in with me (ideally in some form by FRIDAY) if you have not shared progress or a source of data.

Double \(y\)-axes are bad, …

… quadruple \(y\)-axes are better!

Despite the recommendation against double \(y\)-axis plots, parallel coordinate plots exist.

Figure 1: Parallel coordinate plot of Fisher Iris data by Wikipedia user JontyR

Conventionally unconventional

Figure 2: A “spaghetti” parallel coordinate plot for various morphometric measurements in birds (from Getting more out of Graphics.).

How do you feel about this graph?

Paradox

Figure 3: Parallel coordinate plot of Fisher Iris data by Wikipedia user JontyR

Figure 4: A dashboard-generated bar plot.

What do you notice?

Expert advice

A post from Andrew Gelman which links another post, which itself links a paper.

One takeaway: “Be nice to yourself, make good graphs.”

Also: “Experts disagree, experts make mistakes. Do your best.”

Case study in elections

Figure 5: Title page for article DOI: 10.1111/ajps.12004

Figure 6: A three-panel plot relating vote share to income category.

Case study in elections (Maps)

Figure 7: A map whoing changes in vote with respect to demographic category.

Figure 8: A map whoing changes in vote with respect to demographic category.

How to practice? Messy data?

Messy data could be an entire class of its own.

With luck we will explore public data sets later tonight and explore messy data next week in a small-group, hack-a-thon-style “competition”.

I made some fake data in fake_data.csv. Grab it and let’s get to work.