I've got some data. – Now what?

I've got some data.

Now what?

Slides at chagan.github.io/data-talk

You may think your data look like this

But really, it's this

What does that mean

• Come with questions
• Know their bias
  • Who collected this? For what?
  • How sure are they?
• Can't rely only on the data

More likely, your data are this

Quick bath

• Take a min, max, sum and average
• Sort and scan
• Missing values
• Change things around
  • Text to columns
  • Convert to numbers, dates, text as needed
  • Pivot tables

Data smells

• Talk to the people who collect the data
  • Get the documents behind the data
• Check previous years
• Row numbers
  • Excel row limits
  • Round numbers
• Sample size
• Null Island

Make your life easier

• Create a data dictionary
• Make a copy of the original
• Don't make changes in a cell, create a new column
• Track your changes

Visualization

Before you start

• Know your data
• What is the purpose of your visualization?
• What are your users likely to take away?
• Make a sketch (or many)

Types of charts

• Why is this not a bar chart?
• Comparisons
  • Bar charts, line graphs, slope graphs
• General trends
  • Area (bubbles, pies), shades

Design

• Clarify, not simplify
• Limit colors, fonts
• Interactivity
  • Overview first, zoom and filter, then details-on-demand
  • What is the purpose (what will your users to get out of it?)

Tools

• Chartbuilder
• Tableau
• High Charts
• Google Charts

When is a map a map

• The geography is key to the story
  • The interesting trends are tied to the geography
• That story is clear in the presentation
  • Showing those trends on a map is the clearest way to present them
  • Geography doesn't distort the data

Tools

• CartoDB
• Google Fusion Tables
• ESRI Story Maps
• Leaflet (Maptime Tutorial)
• MapBox

Resources

• How to 'interview' a big pile of data
• The Quartz Guide to Bad Data
• When to use maps in data visualisation: a great big guide
• The Functional Art
• ProPublica Guide to bulletproofing your data
• Poynter's Excel for journalists
• IRE and NICAR-L

Lab

• Download Chicago City salaries dataset as csv (Comma Separated Values)
• Load the file into Google Sheets
• Who's the highest/lowest paid employee? Anything intersting about that?
• Use a pivot table to find the department spending the most on salaries. Chart the top 10 in your tool of choice.
• Find one other interesting thing that might be a story. Post that and your chart to this week's slack.
• Stuck? Refer back to "How to 'interview' a big pile of data" on the prevous slide.