Hacker News Reader

Comment by StavrosK | Hacker News Reader

Well, it looks like they vary by -+0.5kg, since this is what I'm doing here: http://www.stavros.io/misc/weight/

That's why I want a smoothing algorithm, but the one I found here[1] is giving me a bit of a problem (I can't see why the data points are being squeezed).

[1] http://www.swharden.com/blog/2008-11-17-linear-data-smoothin...

View on HN

Relative to your overall weight, that doesn't seem like it's highly variable, but I've never been much for stats. Try weighing yourself daily after dinner and see the difference.

How much water do you drink per day? I have found that in times where I'm not drinking enough water, my weight will fluctuate more.

Nah, it's not that variable, it just fluctuates a bit. By "it's not as useful" I mean that weight loss/gain happens at larger timescales, so you can't look at normal daily noise and say "omg im so fat".

I think the noise has more to do about the time you last drank water (right before bed, for example), also the time you last ate (the later you eat, the more food you'll still be digesting when weighing yourself). It's not a problem by any stretch, it's just normal data variance.

you can also use the strategy described in the hacker's diet[1]

[1] http://www.fourmilab.ch/hackdiet/

Worth noting: the author of the Hacker's Diet, John Walker, explains why his choice (exponentially smoothed weighted moving average) is a good choice of smoothing function (vs. some other common functions) for tracking weight over time. See the section "Signal and Noise".

Highly recommended. Beeminder implements it too (inspired by the Hacker's Diet).

Reply on news.ycombinator.com