I finally got around to looking at my MyFitnessPal data that I exported a couple years ago. I wasn’t really counting macros during that time but when I graph percent of protein, carbs and fat calories vs total calories it seems that protein does appear to help curb my general calorie intake, whereas carbs are relatively steady and fats increase my intake. There are a lot of anomalous data points though especially on days where I didn’t log every single food (typically days under 1200 calories) so I don’t know how much I can trust these graphs.
1 Like
All I see is blobs 
Can you do a binned scatterplot, or share the correlation coefficients?
Here is a 3rd order poly fit… Not sure why they all look asymptotic unless mfp has a funky way of calculating total calories (the first graphs I calculated total cals as 4carb+4prot+9*fat)
2 Likes
Yes I guess so… just thought that 700 data points makes for a more convincing argument than 7 lol
Plus I’d rather spend more time walking than sitting in front of my laptop
Calories=f(fibre,fast food visits,sleep,moon phase, temperature,how I feel about my cat, number of emergency calls I receive today…)
Make it multidimensional heatmap. x & y are the macros. Color would be total average calories. no sorry this is not useful.
Edit: Why polynomial? How about comparing Loess and expected linear based on how many calories would be contributed.
Try https://xeno.graphics/multi-class-hexbin/ https://xeno.graphics/scatter-pies/
Or https://data.princeton.edu/stata/graphics #3.3.3 or something like https://stackoverflow.com/questions/36945185/display-color-in-ggplot-2-with-scale-fill-manual-and-scale-fill-discrete-in-stac?rq=1
Actually, just load R and ggraptr and play around with the graphs. That will be relatively fun and easy.
Also could you please give me a sample of the output MFP premium makes?