Twelve Tips for Data-Driven Research

(20-minute talk)
In this 20-minute talk, I cover twelve tips for performing data-driven research, which are relevant for people working across science, engineering, and business fields.

I gave a guest lecture in our department's Intro. to Ph.D. seminar class that all new Ph.D. students take during their first semester. The second half of my lecture is relevant not only to students but more generally to anyone doing data-driven research, in both academia and industry. Here is a series of short videos extracted from that part. The editing is rough since I had to cut out a lot of impromptu discussions to make the videos crisp. Enjoy!

Twelve Sticky Notes

Back when I was a Ph.D. student, my advisor kept on giving me the same tips over and over again for doing data-driven empirical research. I condensed those tips into twelve sticky notes that I pasted onto my office computer monitor:

I stared at those tips every day while I was working, and now I'd like to share them with you ...

  • Tip 1. Start simple
  • Tip 2. Look at raw data
  • Tip 3. Be concrete
  • Tip 4. Incremental results

  • Tip 5. Find something interesting
  • Tip 6. Push hard on clean signals
  • Tip 7. Actionable
  • Tip 8. Comparative advantage is key

  • Tip 9. Find solid levers to pull
  • Tip 10. Document successes and failures
  • Tip 11. Inspect outliers
  • Tip 12. Eliminate sources of bias

Created: 2015-11-05
Last modified: 2015-11-05