Saturday, February 26, 2011

It's a balancing act

So I decided, quite randomly, that I wanted to see how accurate my Wii Balance Board was. You see, we do a lot of test weights of Patrick on the Wii Balance Board and I thought it wouldn't hurt to have some idea what the margin of error was.

(As an aside, the HomeBrew channel is brilliant and adds much value to the Wii. Thank you, Nintendo, for not being ridiculous with your DRM. We can all understand that you don't officially want to support HomeBrew but not actively fighting it is a net positive. We do our weights through the balance board app in HomeBrew.)

Having taken many science courses, I quickly determined that it is effectively impossible to authoritatively prove out the balance board with the tools I have at hand. That is, I don't have any object that I can trust for a true reference weight, and regardless I don't have a number of objects to prove out the accuracy over a range of test weights. My concern, for what it's worth, regarding the range of weights is that my mom used to have an old scale that her 300+ lb husband had stretched the springs on. It was accurate at low weights and not at all at high weights. That is, I believe that there are physical characteristics to a scale that are influenced by the net weight. I don't know if this is accurate given the type of pressure sensors used in the Wii Balance Board but assume so.

The one thing that you can demonstrate is whether the scale is at least consistent.

I did 12 sample weights on an approx 98 kg object, myself, with a standard deviation of 40 grams. Because of the small sample size, my worst sample was two deviations out with most occurring within one deviation.

I then took a barbell weight that was 10 lbs / 4.54 kg and repeated the same experiment. The standard deviation rose slightly to 60 grams, again the worst sample was two deviations out with most occurring within one deviation, and the difference between the two averages was 4.56 kg. I'm not sure how much I want to trust the accuracy of a barbell weight so am willing to accept on the face of it the 20 gram discrepancy in expected weight.

Given that an unfilled diaper is approx 20 g and a filled diaper can be anywhere from 50-80 grams (I base this on the hospital lab scales that we used to measure each of Patrick's diapers when we were in hospital which had an effective accuracy of 1g), I think the error factor on the Wii Balance Board is negligible.

Whenever we've done a test weight at the drs and returned home, it's identical so I suppose in the end that's all that matters. (I don't use Patrick as a reference weight because he can excrete or take in 100 grams between drs and home theoretically.)

No comments: