Applied Nate Silver – now what?

Vacation was not kind to my weight loss plan, far worse than I expected. But once again this raises a big question about how to apply Nate Silver’s techniques to prediction, that is, what should I expect now. So first the facts, or the fact:

I gained 22.1lbs during vacation, or rather between the weigh-ins two weeks apart that included the vacation interval.

Now, as usual this number is discounted two different ways:

  1. the infamous “water weight” which is an illusion, not a “real” gain
  2. a “real” gain of 22.1lbs would require calorie consumption of about 7500 calories/day – no, I didn’t pig out that much. In fact, I actually showed considerable restraint compared to my usual long driving trip consumption. OTOH, it was vacation and we did eat out, but my calculations lead to something like 2000 more calories/day consumption and maybe 700-800 less calories/day burning (only one workout during the trip despite having exercise equipment at most of our lodging). I think that puts my maximum gain more like 10lbs than 22.1lbs.

So if this is all fake it should be easy to discard the non-real gain and that appears to be exactly what has happened. I’ve actually been a bit more restrained (in consumption) and a bit more active (in exercise) so my daily calorie deficit, since returning, has been better than average. And so here’s the result:


This shows an incredibly rapid drop but one that shows signs of decreasing and thus appropriate for fitting with a second-order polynomial which shows a “bottoming-out” in just another day or so (Note: Yesterday’s graph showed I’d bottom-out today and that didn’t happen so the predictive power of this is very limited). Let’s look at the actual changes:

14.41 5.91 2.96 2.77 2.77

14.41 is the total for first four days back from vacation, but you can see that the daily drop, while significantly lower than the first day has, in fact, remained fairly constant.


Need some help here, Nate

(you seem to be ignoring my previous pleas for help, but maybe you’ll respond to this one since it seems like challenging question)

What is the appropriate way to predict with the remainder of the “water weight” is removed, i.e. when do daily drops return to “normal” (and thus much lower value averaging about 0.3lbs/day, not 2.2lbs/day)?

I really felt that today the daily drop would be much less, partly based on yesterday’s graph, but also some intuitive notion that four days of “normal” (i.e. reduced consumption and high exercise) would be enough to counteract the vacation gain. But today’s number provided no clue as to what tomorrow will bring, which again I’d predict as very small drop (possibly even uptick).

As it stands now I’m now at 7.73lbs gain or about 0.5lb/day. That is much more in line with my sense of reduced exercise and increased eating. In fact, since these past four days have been above average for calorie deficits, I might actually believe I’ve lost a “real” 1lb during this time, putting my vacation gain somewhere in the 8-9lb range, which is about what I expected. But of course, “expecting” is another way of saying conformation bias which I know Nate would warn me against.

But relatively soon, certainly by this coming Sunday’s weigh-in I expect the temporary (water weight) effect to be gone and I’ll be back in steady-state. Then I start chipping away at the true weight gain which I think will take at least four weeks to reverse. So all things considered that means a new major data point for me – any backsliding on weight takes 3X longer to reverse than to cause, so that in the future, for maintenance, I definitely need all this statistical stuff working so I can detect a “real” uptick and counteract it in less than a week. In my remaining lifetime there will be other vacations and other excuses for gains, so minimizing those gains and getting them back off will be critical.

This is partly why I’m falling a bit for a recent fad, the 5:2 Fast Diet. It makes sense at some logical level, even though, like all the other fads, it has no science as a base and thus is purely a guess. BUT, this water weight issue is still unresolved (how to account for it in statistics being what is “unresolved”). So if I have an uptick, then do the fast, then see downtick, is that just the water weight again, or did the fast do some good.

The real issue is long-term. It’s still the simple equation. We must consume fewer calories (from any source, type of food, when eaten, all that folklore is irrelevant) than we burn. When we fail to do that we must compensate on future days for the days we backslid (that’s the concept of the 5:2 fast, the fast days will be more extreme deficit than the feast days were surplus, so the sum will be some deficit).

But these huge swings over short periods of time make the data analysis problem huge, so Nate, why don’t you apply your considerable brainpower to this problem.

btw: I’ll be “happy” if the next Sunday weigh-in just hits 200.0 (0.6lbs lower than today), leaving me stuck in the 90s again for the four weeks it will take to burn off the real weight gain. Anything better than that would be delightful but I really don’t expect it.



About dmill96

old fat (but now getting trim and fit) guy, who used to create software in Silicon Valley (almost before it was called that), who used to go backpacking and bicycling and cross-country skiing and now geodashes, drives AWD in Wyoming, takes pictures, and writes long blog posts and does xizquvjyk.
This entry was posted in comment, musing, whine and tagged , . Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s