I have been using my calorie counter I made for a couple months now and I was curious to see how well the number of non-zero calorie items I entered each day correlated with my total calories for the day.
If I aggregate the daily data I get an r-squared of only 0.2 but if I aggregate it by week I get an r-squared of 0.7
I am not very good at statistics but does this suggest that tracking the number of items I eat in a week might be a fairly good predictor of how many calories I ate that week? If so then maybe just counting the number of items I eat per week might be a good enough metric to limit calories. I would assume the monthly item aggregate would be even more accurate estimate/correlate.
I don’t understand if anyone has ever bothered to have patients just count their items instead of calories? This would be way faster and easier and perhaps still pretty effective, especially for people who get discouraged pretty easily at doing a lot of data entry.
Ok, my physicist friend helped me a bit with the data and I realized that I forgot to throw out the zero calorie items in my data plus I threw out the weeks less than 7000 calories since I didn’t log those fully, so r-squared is actually 0.81
I threw the new data into gnumeric and got a pretty low p value
Well I am not sure it is such a strong correlation or perhaps it is a fake one because today I realized that there are some weeks that are missing days that I didn’t log so they would of course have fewer calories that would maybe enhance the appearance of a correlation.
After I collect more data I will throw out any weeks that don’t have a complete set of 7 days data.
I have been tracking food (increasingly adherent) intake since early 2019 when I signed on for a Salk Institute study focused on the circadian clock. The protocol for this Satchin Panda study (operationalized via a custom app) had participants take a picture of what they were eating and enter a very brief description. I think the app entered the time of day. The study didn’t care how much of anything you ate, just what you ate and when.
Thinking back, this practice really improved my health. It lead to wanting to track more precisely, but this is really hard to do given that food are inherently variable, so it’s always going to be +/-.
I use Cronometer, which is good but not without its issues. I have moved to preferentially weighing my food. Lots of obvious variability in volumetric measure. Then, you have to think, even if you get the mass right, there is still variability. The other day I started weighing my coffee instead of recording by volume. The Keurig is off a few grams for each cup of coffee made with the same brand of (organic) pod and mug size setting. I was surprised, but it fit with everything else I’ve observed.
I think this is fun, but I realize this kind of precision probably isn’t very meaningful. I wrote to Panda’s post doc who was running the study to get more info on the underlying theory they were leveraging. Never heard back. Probably time to check for a publication.
I’m marking this as a topic for a virtual QS Conference breakout session, along with the conversation about Personal Data Dashboards. We haven’t figured out how to organize these yet, but we’re working on it. The challenges of food tracking are very important and a lot of people in the QS community have learned useful tricks; but clearly the problem remains unsolved. #virtualbreakout
I think there can be some benefit in having people discover how calorie-rich certain foods are, but that works even if any data that is logged is discarded…
@mookiebearapps I like your approach of trying to identify an easier metric to track than calorie/macro counting. My personal belief is that calorie counting is a fool’s game - without meticulous measuring and weighing there’s too much variability. And non-chain restaurant meals are really just a crap-shoot guessing game.
I believe that one day, we’ll be able to simply snap a photo of our food and software will be able to accurately calculate ingredients, portion size, macros and calories but I don’t think that’s a reality at the moment.
But I wonder if simply counting the number of items is too simplistic to catch meaningful variations in the quality of food consumed over time and their affects?
Counting items may correlate somwhat with total calorie consumption, but it is incidental; i.e. it is not the number of items that is actually driving weight loss or gain, but the number of calories consumed. So my question would be, why count some incidentally correlated variable when you can quite easily just count the causal variable itself?
Counting calories seems far from easy to me, and way too tedious to do for extended periods of time!
I suspect that for most people, the main benefit of this exercise lies in the increased awareness of their eating habits; but this can also be accomplished with logging simple meal summaries, or even just taking photos.
