Friday, January 14, 2011

Fake data labs

One thing I've noticed over the years of teaching upper-level physics labs is that the students see the experiment "working" as the finish line.  They seem to plan their semesters with that as a goal, determining when to order equipment, when to test it, when to set it up etc.  And often it works out, the experiment "works" right at the end of the semester and they're happy.  Of course, I'm usually not because there's not enough time for them to do a decent amount of data analysis and write-up.  Don't get me wrong, I think the experience they get by having to plan and execute their own experiments is good, I just needed to find a way for them to experience the other stuff too.

My solution, which I've used for the last several years, is to have them do a fake experiment.  This is in our Modern Physics lab so they're sophomores typically.  I have them choose from a list of seminal modern physics experiments, things like the photoelectric effect, Compton scattering, Michelson/Morley etc.  I then ask them to plan a 21st century version of the experiment and to assume that money is no object, since they're not going to do the experiment in real life anyway.  I have them plan a fully computer-controlled experiment (we teach LabVIEW in this course) and ultimately they have to turn in a functioning LabVIEW program, a formal lab report, and a Mathematica document showing their data analysis.

Now this last part is where I come in.  About half-way through the semester they have to turn in to me what their raw data would look like.  Typically this is a description of a data file or files explaining what all the columns and rows would be, as saved by their LabVIEW program.  They have to figure out things like the motor step size they'll use, what wavelengths of light are necessary etc. I then spend spring break creating fake data for them.  I do this in Mathematica with a random number generator providing appropriate levels of noise.  I also choose my own values of things like Planck's constant and the speed of light so that I can know that they've done their data analysis correctly.  I especially like choosing a direction for and relative speed of the luminiferous ether.

What's great about this project is that the students always have time to do the data analysis and write up because they don't spend any time on the actual experiment.  Of course if this is all we did it would be a problem but they get a lot of experience with real experiments both in and out of this class.  Being somewhat early in the program students get a sense of the type of planning they'll have to do for future long-term projects and they get a real sense of the difference between raw data and a result.

I always tell the students that their last sentence in the report should be something like "Planck's constant is ___ +/- ____".  I tell them that's their goal, not getting the experiment to work.

2 comments:

  1. This is awesome. I love that you get to play with the dials of the universe and have them figure out what changes you made. If you have a chance, I would love to see a sample report submitted by some students. I'm also very curious as to how I could bring something like this down to 9th grade. I have on occasion given them excel files with fake data for acceleration vs force of a cart, and had them deduce the frictional force on the cart. I also have given out fake data for a pendulum on the moon (with over 200 combinations of length, mass and period), and had them "search" the data figure out the strength of gravitational field on the moon.

    ReplyDelete
  2. Here's an example of a doppler shift/galaxy speed write up from last year:
    http://bit.ly/hp17e1

    I like your approach for the moon's gravity. The reason I make stuff up is because I'm tired of what I call BOTBJ (back of the book justification) where students do amazing math tricks to ensure they get the "right" answer.

    ReplyDelete