Applied Nate Silver – slowdown

It’s been two weeks since I reported my progress on weight loss and so here’s the update. I’ve definitely slowed down, both doing less calorie burn in exercise (partly due to doing real biking and strength training which have no calorie burn data) and a bit less well on dieting (experimenting with the 5:2 (feast/fast) diet plan). With five Sunday weigh-ins (since vacation blip) my weekly loss rate has dropped to 1.35lbs/week (vs 2.39/week for first 24 weeks (pre-vacation)). But some of the slowdown is simply due to the fact that I’m approaching my long-term target (which decreases my urgency), and I’m attempting to re-distribute weight (more muscle via diet changes and different exercise, less fat), and I still have another blip to go (Wyoming trip) before reaching “steady-state” (probably around August). So here are some of the data, including some new types of analysis:

First, the “money graph” (my “bottom-line” long-term trend):

weekly32-money

The blue markers are the pre-vacation, relatively consistent “aggressive” loss, the red are the vacation “blip”, and the green are the “back-on-plan” (albeit less aggressive) of recent weeks (the next graph shows this even better). The recent data at least beats the first 24 weeks even though it has clearly slowed down, as seen by the two different trendlines. But note the bottom of the vertical scale is my “target” so I’m getting close. The combination of the vacation blip and the recent slowdown around nine weeks behind the original predictions, but in this case I’m OK with that since I’m now more focusing on final body shaping and achieving steady-state, the real (and most difficult goal) of all this exercise.

weekly32-longterm

With the much lower weight as well as slower rate of loss the variability, both daily and weekly and scale variability is definitely now creating “noise” far larger than the signal (real long-term trend, as seen in the daily blue dots in the graph above). And it feels subjectively, as well as analytically, that the noise is getting even larger. What I find subjectively is that daily weight is now almost completely uncorrelated with calorie deficit and instead correlates almost entirely with previous day volume of food (and esp. salt, I believe, but can’t prove) because more food (salt?) means more gain of “water weight” (whatever that really is, i.e. restored glycogen and its hydration?).

So let’s try to analyze real weight variability (as opposed to measurement variability). Here’s a somewhat hard-to-read, but interesting graph:

weekly32-weightvariability

The blue markers are the average of daily weigh-ins; note that due to other experiments each average is NOT based on the same number of scale samples so the blue points are not completely self-consistent in definition. Nonetheless they are plotted and the linear regression trendline is shown. Visually it is easy to see that more recent “weights” have more variability (scattered on the graph) so the red markers (with the rightmost vertical axis representing the grid) show this. The red markers are the measured weight (daily average) – computed (from the trendline, computed for that day) weight, or about the best analysis I can think of to represent the daily variation. Again visually it is possible to see that the daily variation itself seems to have great range in more recent data. So let’s just isolate this:

weekly32-weightvariability1

This bar graph, with red bars representing Δ above the trendline (bad, so shown in red) and the green bars representing Δ below the trendline (good, although sometimes just as “wrong” as the upticks). Clearly the magnitude of the variation has increased in the recent days. Now, again, given my reduced focus on daily progress and various experiments (both with data and with actual exercise/eating) the greater variation is not unexpected (although not desirable). But the real point is that it is getting much harder to detect signal (real longer-term weight loss) due to this high level of noise.

Now what does this mean?

  1. It does show how relatively difficult it is going to be to monitor “steady-state” weight, that is, to spot real deviations (like actually regaining real weight = fat) vs just daily variations as my life becomes more “normal” (some big eating days, some low exercise days). Getting a handle on this and having the analytical tools to monitor “reality” will be critical to success in weight maintenance, the hardest part of fitness, for both me and other people. The odds are not good – most people (like nearly 100% in 3-5 years) regain all their lost weight after a concerted effort to lose weight. In my case I must succeed because this kind of aggressive weight loss I’ve done in the past eight month will probably be impossible for me later in life; IOW, this is IT!, this is the loss I have to maintain for the rest of my life.
  2. All the data gathering and analytics, as represented by all my many posts in this series, are probably more extensive (more OCD-ishly monitored than most people do). Given the variability I’m seeing I even wonder how other people can monitor their weight and also why diets (esp. fad diets) appear to work, albeit briefly. But how can even more people with more modest goals, trimming enough fat to fit in recently old clothes, where the weight loss plan might be like 10lbs in 3 months, accurately monitor their progress. In short, is most weight loss the victim of this high signal-to-noise problem, which would then suggest, as an aid to people trying to control weight, that they need much better tools than the typical diet book or website (a business opportunity?)

Hey, Nate, figure this out and publish a new book

So I expect to bounce around during the summer, at least one actual gain (my trip), some short-term fluctuations, fails in experiments (like 5:2 diet, does it really work after significant weight loss or just early in the program). But most of all how will all my obsession with data and analysis turn into practical tools I can use for the rest of life to maintain a “healthy” weight and not just reverse this hard-won progress.

Prediction:

185.0

by 4th of July. Check back then for yet more long posts.

Advertisements

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, prediction and tagged , , . Bookmark the permalink.

2 Responses to Applied Nate Silver – slowdown

  1. Thank you for the auspicious writeup. It actually used to be
    a leisure account it. Glance complicated to far delivered agreeable from you!

    However, how can we keep in touch?

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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 )

Google+ photo

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

Connecting to %s