Study Design: the best model for a cat is... a cat!

Accepted Session
Short Form
Scheduled: Tuesday, June 18, 2013 from 10:00 – 10:45am in B301


With good study design you can state how confident you are that you have a cat. You can hypothesis test your cat--is my cat like other cats or is it a dog? You can even design an experiment to determine the correct feeding time for your cat.


Good study design is like test driven development. In TDD you write the tests before development. In study design you clarify and define the problem or question you want answered before you start collecting data. Pilot studies are performed to see if your idea is workable, or if you need to make adjustments, just like we test code to see if it breaks on Y2K.

Data is everywhere and easily collected. This talk will cover the fundamental statistical principles that allow you answer questions and make predictions. We’ll talk about barriers to effective experimentation, like poorly defined logic. (I was once asked to study a medical problem where the definition was “I know it when I see it.") We’ll also look at some basic R code for frequencies, distributions and histograms to help you analyze your data once you’ve collected it.

Here’s why this is important to me. At times I have given up and pulled my hair out over bad, insufficient study designs. Good planning up front would saved my clients so much money. I have done many research studies. I had to defend studies when the conclusions conflict with conventional wisdom. Solid study design has backed me up, and it will do the same for you.

Speaking experience

Bar Camp, Ignite and a Thespian