Applied Nate Silver – Disaster! Not Believable – 2

More daily data is now showing me a new variability I hadn’t (amazingly) spotted before. In all this analysis I was well aware of scale variability but my early data, a very steady weekly decline, lulled me into believing there was no significant temporal variation. Now the surprise from last week and now having collected six days of weekly data shows this is a much bigger factor than I had imagined (actually hadn’t imagined it at all, except as small random fluctuation on top of a generally linear trend):


The trendline is based on the red points which are the weekly averages (weeks 14-18) and the latest daily average (today week 19+6/7) and ignoring week 19, the highly anomalous point that started this entire thread. Today’s weigh-in, which as you can see is now back on the long-term trendline (which fits all the points very closely, as you can see and is proven with the very high r^2).

But the interesting part is not just the week 19 anomalous point but also the daily points thereafter that were nicely trending down, but now all appear to be substantially off the trend. Or, put another way, the extreme calorie deficit I did this week (the most ever) seems to have made up the difference (at last) for the anomalous uptick.

But to show how crazy these data are (compared to underlying calorie balance) the drop was 7.7lbs in just six days! OTOH, my recorded calories_burned – calories_consumed was 9704 which accounts for only a 2.8lb drop, leaving 4.9lbs unexplained. Given I dismissed the anomalous week 19 point as “water weight” (due to the high consumption of soy) it’s hard to explain: a) the large magnitude, b) the time it took (six days) to finally dissipate, and, c) the unusual drop today.

So, the lessons learned in this exercise is that I do potentially need to monitor daily weights despite the fluctuations I’m going to see but then use only (non-anomalous) weekly weigh-ins for the long-term evaluation (as most weight loss plans recommend). The daily data will (perhaps?) explain any otherwise unexplained anomalous weekly values.

Now the perfect exercise to close this out would be to see what happens tomorrow. It seems likely today is another anomaly and thus tomorrow I might see an uptick. Certainly this graph would tend to predict that:


Just based on “eyeball”/intuitive prediction today’s values seem too low and therefore the weekly trendline (rather extreme rate of 5.5lbs/week loss, or 2750 calorie deficit, which is just barely imaginable based on extreme restraint on eating + normal high exercise (exercise calories have exceeded consumption every day)) would tend to predict about the same distribution of weights tomorrow (and thus average about the same as today) as occurred with the points of Monday and Tuesday (essentially no change).

That’s what the math might predict but I’m guessing Nate would come up with a different answer. But here’s one more piece of the puzzle. Today is a family celebration where I’m making break and grilled pizza and lots of other food (including scrumptious desserts from my wife’s sister) will show up. While I can resist eating this morning + probably have time for good exercise calorie burns there is no way I’ll restrain myself later tonight.

So I’m going to actually make a very pessimistic prediction (at least in comparison to the data) and say:


which is 1.3lbs more than today, but I think today’s value is anomalously too low. That would also give me a two-week calibration of average drop of 2.0lbs/week drop which was about my most recent rate.

So we’ll see.


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.

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