Discovery Summit can be a lot like a choose your own adventure book.
Hannah Fry's presentationDelegates choose among many amazing sessions, leading them to new knowledge or interests that might inform their next choice.
This week, they also had the opportunity to customize a plenary session. Professor Hannah Fry of University College London gave a highly engaging talk featuring a variety of her favorite data stories. She invited the audience to vote on topics based on their interests.
This audience chose to hear about naughty AI, serial killers and cows. (Martin Demel noted that cows are a most appropriate topic because that’s how SAS got its start! @martindemel). Many of Hannah’s examples provided a cautionary tale about what can go wrong when we:
- Lose the context around an analysis project.
- Gather a ton of data without thinking about the question at hand.
- Use data that is readily available, but not ideal for the problem set.
“I especially liked the point she drove home from Daniel Kahneman about substituting an easier-to-answer question for a hard question—a proxy, which isn’t the same,” said Anne Milley (@anne_milley), who facilitated the session.
While it’s not easy to have a conversation about serial killers, Hannah provided examples to demonstrate how data can assist police in their search – which prompted some intriguing audience questions. She also shared how this methodology is being applied to other worthy causes like locating stagnant water that causes malaria outbreaks, finding bomb factories that make IEDs and uncovering Banksy’s identity based on the location of various paintings.
Hannah Fry's first example with data from the Challenger disaster
Exploring an affinity group for women in science and engineeringLooking for more inspiration? Hannah recommended the following books:
Calling Bullshit, Carl Bergstrom and Jevin West
It All Adds Up, Mickaël Launay
The Hidden Half, Michael Blastland
The Art of Statistics, David Spiegelhalter
Any of Tim Harford’s books
Bright Minds Mingle
Delegates had more choices to make during the Bright Minds Mingle session. We overheard conversations about:
- How to foster inclusion and elevate women in science and engineering.
- Tips and tricks for getting management support for analytics projects.
- How to build a vibrant JMP Users Group.
- A demo on how to simplify the automation of data.
- And a healthy debate about the true definition of “pie.” (For example, must pie have a top crust? Is pie savory or sweet? Does whipped cream count as a top?)
John Sall plenary session
On day two of Discovery, John Sall warned us of eight things to worry about along the modeling journey, providing a mnemonic device to jog our memories: MOONSHOT.
- Missing values
- Outliers
- Overfitting
- Non-normal distributions
- Selection bias – multiplicity
- Hidden effects (Simpson’s Paradox, aggregation bias)
- Overlooking split-plot structure
- Truncated data, limit of detection
Paper and poster presentations
A pre-Pi Day celebration
A pie-baking demonstration
Jill Santa-Lucia from Catering Works
The adventure continues...
We’ll reconvene March 22 and 24 for more. It’s not too late for customers to register, so don’t hesitate to invite people who may reap the benefits of this event.
