## Applied Nate Silver – another (positive) surprise

All week long I kept feeling I was eating too much and exercising too little but it appears I was being unduly pessimistic. Which also means the analytic tools, both Nate Silver stats and the BodyMedia analyzer are predicting more accurately than I am. So I guess the answer is: believe the data not my gut. So here’s the money graph:

Note: I went back to linear trendline since third-order polynomial was only slighter different. Furthermore the long-term slope is showing the difference in weekly change vs the earliest, more optimistic data (Δ=2.7).

The predictions I beat were both the predictions in last week’s post, but also the methods (and then average) I did last night of the BodyMedia data:

1. actual Δ = 3.1
2. (Σcalories_burned – Σcalories_consumed)/3600 = 2.9
3. #2 discounted to 2/3rds (previous discount) = 2.0
4. ((Σcalories_burned)*.9 – (Σcalories_consumed)*1.15)/3600 = 1.8
5. based on 2146cals/day as sedentary requirements, but for 20 hours (since average 4 hours of exercise) = 1788 and then 1197 cals/exerciseday = 2985cals/day (BodyMedia measures 3216) and calorie intake = 2.5
6. average of raw BodyMedia and my old estimating technique = 2.7
7. my prediction last week = 1.8

So all my techniques are now running pessimistic (under-estimating loss) so I should probably get a little more positive.

Now for next week I am planning fancy dinner for Valentines, so that blows out that day’s calorie intake by probably 1000 and my exercise this week was really tough so I expect less and so the net result of my “intuitive” guess is a Δ of 2.0 (well below by long-term average of 2.5, but I also think today’s weigh-in might be a little low and therefore some of the delta next week will just be accommodating that), so my prediction is:

# 205.7

Let’s hope I make that since I’d also love to have broken 200 by week 20 where as little as two Sundays ago I was predicting only 205.

I also am going to have to consider, against normal advice, self-adjusting my meds because I’m getting increased (and more severe) episodes of light-headedness, which I normally attribute to hypoglycemic state but probably now is more likely due to low BP. So I measured this morning and am now 100/52 which is the lowest I’ve ever seen (my rest HR is running in low 50s). What I interpret for all this is that my push in the last three weeks to do some of my exercise sessions at the maximum exertion I can do is working and pushing my CV levels to nearly athletic levels. Hence taking a full dose of BP meds is excessive, so I think I’ll try the pill splitter and reducing the dose and then carefully checking both HR and BP during the week.

So potentially this weight loss and fitness is accomplishing what my doc said 15 years ago – get in shape or take meds, so maybe now I can drop the meds.

p.s. Here’s another extra little bit of analysis, esp. as I now am getting enough data where this may make some sense:

This is the raw weekly changes. As you can see this is a fairly random pattern and the r^2 shows that, but I choose to believe the slight negative slope in the trendline (meaning my weekly losses are getting lower, which every other analysis confirms and this also matches basic principles). But what’s interesting is not the trend but the higher volatility of the data (since the holidays broke my relatively consistent weekly results) which probably explains my “gut” feel of making less progress.

p.p.s. Couldn’t resist, here’s another graphical tidbit of my process now that I have four weeks of data in BodyMedia:

Every day is a deficit, which is really good, and most exceed the “plan” I set up (2lbs/week losses). I know the BodyMedia data has some definite flaws, but interesting it has a tendency for under-estimates (of calories burned) to be offset by over-estimates so the total isn’t too far off my other data.

This graph is the one that will be really important when I finally get done with the loss and then start trying the hardest part, maintenance, which is something almost everyone (including me in the past) fails at.