Monday, July 13, 2015

A more interesting way to learn experimental design

My colleague Rob and I were talking last week about how he used to introduce the big ideas of stats.  He told the story about the farmer (whom I suppose has a name, this is a true story) a few centuries back who wanted to understand cause and effect with his crops.  Since the problem was mixing together all the input variables, he broke his land into plots and tested each variable separately.  This was (as far as I know) the start of modern experimental design.  I loved the story but didn't think most students would buy-in.  They had to feel the experience of the farmer somehow.

Since I love simulations for everything, especially in stats, I thought a simple farm simulation with a few plots and different parameters you could set would be a good starting point.  The idea was to build a simulator with too many variables (inputs) and a poor understanding of detailed cause and effect.  Students would work in groups of 2 or 3 and have 10 "growing seasons" to learn all they could about how the simulation worked.  Then each group would have one shot to try their inputs on my screen and advance in a tournament bracket (everyone loves a good bracket) to the final farming round.  The most profitable team wins.

With only 10 rounds to learn from, only teams with a disciplined approach would be able to learn enough to give them a strong shot at winning.  Haphazard clicking would leave a lot up to luck.

The first time through, I would introduce the game before I taught them anything, using the simulation as an intro event.  After playing and finding the winners, we would talk through their approaches during the 10 practice rounds and what they were able to learn.  We would discuss the challenges they found in trying to learn things.  THEN, I would introduce some of the core concepts of experimental design.  After some teaching, I would throw them back into the simulator (with a different crop / parameters) to see if they are able to better learn the ideal conditions and everyone can be more profitable.  I would assess them on how they approached the second game (they could write a summary of what they did, how it compared to the first time through, and why they made changes).

Here is an ugly but functional prototype to get the idea across.  I would love it if you can play around with the farm simulator and my description above on the lesson intent and offer feedback / ideas / links / anything to help me make this better.  Both the lesson and the simulator are very rough at the moment, but I think they could be awesome if it became a community-designed lesson.  Of course all of it will be shared freely if you want to use it.  Thanks in advance!

1 comment:

  1. It was fun to play around with! I think it would be cool if it included graphs/charts showing past input choices and the results that came with those choices.