σ = (X1-Xavg)^2 + (X2-Xavg)^2 ...
√ -------------------------------------------------
N-1
? Standard deviation? Damn it's been a really long time since I've done stats...
SO first we'd have to average weight change (Xbar), then calc our std. dev., then if we were going to use a confidence interval to find percentage +/- bias,
CI: T(predefined value based on sample size) * Standard Deviation
--------------------------------------------------------------------------------
[SqRt] N
That'd be kind of cool. But there'd be all sorts of reporting problems (e.g., females underreporting, males overreporting, etc).