It’s been almost a month since last postings so I think you can stand a few more of these. Just a quick summary of the last six months. I hit an all-time (adult) low about six months ago and was more or less at my target weight and ready to try to do maintenance after eight months of steady loss. Then I got hit with my medical “crisis” where I decided, on doctor’s advice, to put back on some weight in anticipation of treatments; plus I didn’t really care about weight as a top priority any more. That new regimen was followed by a good trip to Texas with plenty of indulging of tasty food treats and so a regain of about 25lbs. And as of today that’s now gone (again) and in fact on my official weigh-in day (Sunday, but if you’ve read other posts you know I have loads of daily statistical day for analysis) I did hit a slightly lower all-time weigh-in day weight, IOW, my six month roller-coaster ride is over. The following graph shows all this history.
And in finer resolution just the last six months “roller-coaster”.
In the top graph (with “official” weigh-ins + a few interpolated points) the blue markers are the 73 weeks of weigh-ins and the purple line is the eight-week moving average (still not quite back to previous low) and the blue line is the 6th order polynomial fit, which shows the hazard of extrapolation as its most recent direction will be all wrong (I plan no further reduction, just maintenance, check back in a couple of months to see my results). The brown markers show an interesting pattern, my average (each dot an average back to week 1) weight over time. It will now slowly decrease and over the entire 73 weeks shows what the load on my heart has been.
The higher resolution chart shows the daily detail (blue markers are the actual weights with the 0.7 standard deviation in scale readings and the red markers are the daily average, usually from five scale samples). This higher resolution shows the much greater volatility of daily changes, which I’ll detail a bit later. The downward sloping linear regression line shows by reloss at about 1.2lbs/week, which is substantially less than my initial very aggressive loss of about 2.3lbs/week. The upward sloping regression line is almost meaningless but represents the long-term trend over this “roller coast” which is still positive and will take a few more months of maintenance at my current weight before it goes flat.
But let’s looks at a few more interesting details.
The blue markers are the daily weights (usually an average of five scale samples) and the blue dotted line is the regression fit of these (0.176lbs/day loss). The cyan line is 7-day moving average which smooths some of the volatility thus better showing the actual weight over time (with the several upward blips due to pigging out == having fun eating). Normalizing the individual daily weights to the trendline shows the daily volatility (below).
You can see there are some pretty extreme variations (these are the change in one day to the next), a fairly amazing maximum uptick of 6.47lbs! and maximum downtick of 3.91lbs, and overall, 45 upticks and 59 downticks, with an average of -0.21lbs/day (matching the longer term trendline results). The gold line is the 7-day moving average. So even though today’s weigh-in is a new record low my moving average is just barely down to neutral. IOW, I’ve had periods where I beat the long-term (as measured by regression line) trend (the big blusters of green downticks) and then a few falling-off-the-wagon pigouts (the clusters of red upticks). But here’s the more interesting part of that.
The chart above shows the histogram (with a 3 point smoothed line) of daily variations. Not surprisingly it looks more-or-less like a Gaussian distribution centered on little change, but the tails on both sides show how much volatility there is in daily changes even when there is a long-term overall decline. The standard deviation of these changes is 1.7lbs/day and the range is 2.0lbs. IOW, the key observation is that daily volatility is larger in magnitude than my average weekly! loss.
The fitness freaks (mostly) say to ignore the scale, especially daily values. I think this is exactly wrong. The real thing to do is get more data (including multiple samples per day to deal with scale imprecision) and do these kinds of analysis. Daily changes do matter! And for me “bad” days trigger me to redouble my resolve to reverse those gains. The long-term changes don’t come automatically, it takes effort. And knowing exactly what is going on is critical. Knowing there is a lot of variability in daily numbers means you don’t have to freak out (or be delighted) in one day changes BUT it does mean you have to pay attention. Just because a lot of data like this shows it is “normal” to shoot up on some days, the negatives have to offset the positives. In looking at the daily change bar graph (the red and green bars) you can see it’s much easier to have a “bad” day (uptick) than “good” days (downticks) and, at least for me, the downticks are never quite as large as the upticks. So it takes more downticks than upticks just to average a flat value, and significantly more downticks to register a loss.
Ignoring volatility creates a misleading weight loss plan, so more data (but properly analyzed) is better than less data. So fitness freaks, you’re WRONG about this!
But one final analysis I can extract out of all this is the weekly pattern. Most time series data has some “periodicity” in it as days of the week (or seasons of the year, so sometimes known as seasonal adjustment) change your lifestyle and that has a statistical effect. So here’s the pattern of daily changes by day of the week.
Here the trendline shows the critical part of the analysis. You can see each day (day of the week, 1=Monday) has a wide range, but the trendline (despite the low r^2) shows there is a decline (over many weeks of samples) each day of 0.17lbs. IOW, Monday is my worst day and Sunday is my best. Now this fits “first principles” for my lifestyle. On Sundays, here at Starbucks, I allow myself a few treats plus a food meal (and that horrible stuff, “junk” food) and so naturally I’ll have an uptick on Mondays (in fact this is mostly “water weight”, not new fat). Then I’ll get serious about exercise (Sunday is a rest day for me) and reducing calorie intake. If I don’t get back on track by about Wednesday then I throw in a fast day and a very high exercise day (as it is I’m averaging over 1000 calories/day burned in exercise). And so most of the time by Sunday, for my official weigh-in day (and having a particular “official” weight per week is critical, IMO, to accomplish long term change) I’ve managed to reverse the previous Sunday’s “fun” and get back on track.
Now that I’m back to target weight all this will be critical for me to know and understand and then follow in order to achieve maintenance. During the period of weight loss it’s easy, usually the scale is “good” news. But once you’ve accomplished the weight loss (which still less than half people on fitness kicks do) the hard part is maintenance because you no longer have that “reward” over steadily lower weights. Restricting calories and doing lots of exercise just to stay at the same weight is much harder than losing weight, and certainly harder than gaining it (that, unfortunately, is all too easy and 99% of people fail this).
So my point of all this post is data matters! Probably most people don’t get all this data because they don’t know what to do with it all. But now, having using the flexible but relatively crude tool (Excel) I think I could now build a much better analysis program to encourage people to get lots of data and then understand what it means in order to motivate them to accomplish their goals. People would not have to understand Nate Silver level of statistical understanding to use a tool that already has all these various analysis built into it. So the fitness freaks should be encouraging the development and use of such a tool, plus using data, not ignoring it. Sure raw data can lead to inappropriate actions, but proper use of data could help a great deal, for most people, I believe, in their success.
So, scolds, stop blaming big food or addictive/toxic substances and just admit and explain the real program people need to do weight control. Nate Silver can help people a lot more than Dr. Katz and Mark Bittman and their magic superfoods and nonsense new-agey mindfulness crap.