Thursday, February 22, 2018

Teach 180: Smelling Parkinson's (Day 110)

Today students were introduced to hypothesis testing with an activity developed by Doug Tyson.  The lesson begins by showing students a video about Joy Milne, the woman who can smell Parkinson's disease.  (If you don't like this particular video, there are many others online.)  Students learn that Joy smelled t-shirts and correctly identified the shirt as being from a Parkinson's patient or not in 11 of 12 t-shirts.  We then wonder how likely it is to get 11 of 12 identifications correct.

We set up the null and alternative hypotheses (Ho: p = 0.50 and Ha: p > .5) and then did a simulation by hand to determine how likely it is for 11 of 12 identifications to be correct by chance alone.  The pictures below show students smelling index cards that have either a P or NP on the back of the card.  (It is quite fascinating how excited students get when the guess correctly.)

We plotted the results of 40 simulations on a dotplot and got an estimated p-value of 0.025.  (One student got 11 of 12 correct by guessing.)  Finally, we did 10,000 runs of the simulation using  statkey.  Based on these simulated results, our p-value was 0.0045.   This simulated p-value was definitely small enough for us to reject the null hypothesis.

Incidentally, one of the patients that Joy had identified as not having Parkinson's returned to the lab several months later to report that he, in fact, had Parkinson's.  Joy was actually 100% accurate in all of her identifications!!

No comments:

Post a Comment