Nutrition Works for the AwkEng [warning: discusses diet / weight loss]

Hi all!

[ content warning - today's post talks about weight loss and diet ]

Today's post is about correlating real world effects with data, and more specifically, with nutrition and weight loss. I've dropped about 10% of my body weight in the last 5-6 months, which puts me somewhere close to what I call "pre-dad bod", which I'll count in the wins column.

I'm not here to dispense medical advice or endorse a program, but I find it incredibly satisfying when real world evidence and physical phenomenon are manifested in actual data, and that's what I want to focus on sharing today.


Here's a quick summary of the program I was in

  • First, I was working with a nutritionist throughout the program.
  • I was meticulously tracking everything I ate the entire time. This means weighing it, looking it up in an app, and recording the calorie intake/macronutrients. I also weighed myself every day.
  • So the overall effect of the program was to run at a net calorie loss.
  • We never got so super sciency as to directly measure my metabolic rate / calorie expenditure, which is possible with machines that measure oxygen consumption. Instead, what we did was estimate it based on a few factors like height, weight, age, activity, etc. Again, it's an estimate, not a measurement, but in true engineering fashion, you can get close enough.

The Overall Results

While it varied, I lost on average, about 1 lb per week for the past 20 weeks.


Process Noise and Measurement Noise

Besides the overall trend, one of the things that jumps out to me is the noise in the graph and all the little fluctuations.

There's a lot going on there, include

  • quantization error - my scale only reads in .5 lb increments
  • process error - how many calories did I really eat? how many did I really burn?
  • measurement error - how accurate and precise is the scale, really? I swear I could see .5-1 lb changes just stepping off it and then back on again. And at various points, I traveled to visit family, and used a different scale, as well.

I would always weigh myself and record it when I got out of bed in the morning and it was typical to see 1 lb fluctuations from day to day, although as high as 2 or 3 lbs wasn't unseen. And because I was curious, I would occasionally also weigh myself (although didn't record it) at various points throughout the day, and could see shifts as high as 4-5 lbs from morning to evening, depending on my hydration level and food/water intake.

Sometimes It's Not Process Noise or Measurement Noise

Later in the program, as I became more in tune with my body, I could correlate what I ate with fluctuations in my daily weigh in. Certain foods might make me feel bloated and highly salty foods would make me retain water. I could then see it on the scale. It wasn't noise in the signal, it was real, and I could feel it.

Filtering Signal from Noise

Of course, when there is noise, there are several engineering approaches for dealing with it. A low-pass filter certainly comes to mind, where you ignore the day to day fluctuations and focus on the underlying, slower moving trend.

People love immediate feedback, however, and another engineering approach is to use sensor fusion to combine data from several noisy signals to get a better sense of where you truly are.

In this case, a combination of signals include diet (what I actually ate and drank), clothing fit, waist measurement, physical appearance, self assessment, and activity are combined with the weight signal to get a proxy for overall health, which is the real goal. Actual scale weight is one component of that, it's just the one that's easiest to measure.

Final Thoughts

Ultimately, there were a lot of factors that let me control the inputs, meaning my activity level and what I ate. I know that's the whole crux of it right there. The underlying science though, that net calorie deficit leads to weight loss, absolutely works and it shows in the data.

best regards,

aka THE Awkward Engineer

Leave a comment

Please note, comments must be approved before they are published