As previously covered I've tracked my diet and various health metrics consistently over the last two years, with some sporadic data from before then as well, I export all of the Cronometer data to my Excel sheet, as well as the various Fitbit and qualitative lifestyle factors that I track. As mentioned before I'm not particularly worried about the absolute value of some of the Fitbit markers but rather their trend or correlation with other things, based on my experience for myself the Fitbit HRV for example has been quite a reliable marker of my overall recovery and wellbeing.
From here each day is organized as a row of data with the various columns dedicated to each nutrient or health metric.
Here are all the correlations as I normally view them right now.
A lot of data and information! Don't jump to any conclusions yet! The top row shows how many data points are in that column, followed by which metric is in the column, and then I have my correlation values which are calculated based on that metrics correlation to the health metric that I'm interested in such as RHR or HRV, followed by the name of the metric again and then the maximum, average, and minimum values of that metric. So for example; the first column tracks calorie consumption, and here shows a negative correlation with the next day HRV with an RSQ (R^2) value of 0.059, a magnitude of 29%, and a significance of 1.7%.
The magnitude is calculated from the slope of the correlation, multiplied by the maximum variation/difference of the metric in question (Calories), and converted to a percentage of the maximum variation of the metric I'm analyzing (HRV), and the significance is the magnitude multiplied by the RSQ. Each row is color coded using the automatic excel cell coloring feature with the top 10% and bottom 10% being red or green, and values close to zero being clear. (slope cells not colored as I usually keep it hidden).
Essentially this shows that my calorie consumption has a maximum explanatory power of 29% for my next day HRV, however, the correlation is quite weak with the low RSQ of only 0.059 leading to a "significance" of 1.7%. Now I do realize this is not necessarily the statisical approach used, however it works for my purposes and allows me to quickly see which nutrients or lifestyle factors might need a further look.
So on that note; how does HRV and calorie, and carbohydrate consumption correlate? Using my data from cronometer (which excludes a few of the data points from older data before I used cronometer) we can see the RSQ increases slightly.
There does appear to be a correlation between total calorie consumption and next day HRV from looking at the graph, and in my experience my highest HRV days are almost always while I'm eating at maintenance or in a slight deficit. However recently I've started to question the strength of this correlation, as I've tested finishing eating my food earlier in the day, and this seems to significantly help, oftentimes I would've in the past eaten my last large meal before bed around 10-11 pm and sleep around midnight, however since moving my last meal-time to finish between 7 or 9 pm, I've been able to eat a higher amount without having such a reduction in HRV.. This has only been tried recently, and I'm currently in a weight reduction phase so it will be some time before I can test my theory out properly. For reference I usually have breakfast around 10-11 am.
Lets move to a stronger metric, and one that I find interesting as well, Net carbohydrates-from my cronometer tracking.
Net carbs have a stronger correlation even than the total calories consumed, which is interesting given that my total calorie consumption is highly correlated to my carbohydrate consumption, as when I eat more it is almost always in the form of carbohydrates.
This might indicate that carbohydrates seem to have a more negative effect on my HRV than the other macros. Interestingly fructose and glucose seem to have no correlation whatsoever with HRV, and it is mostly driven by starch and added sugars in the form of sucrose. This could be due to the endotoxin effect when eating significant amounts of carbohydrates leading to gut flora growth and subsequent death.
A commonly known strong factor that suppresses HRV is alcohol consumption, and this is verified in my data where alcohol consumption has a magnitude of 43%, and a significance of 3.2%.
Now despite the RSQ of 0.07, I do believe the effect of alcohol on HRV is very significant. Clearly, there is a large spread for the days where I do not consume alcohol due to things such as activity and being sick (the strongest factor in poor HRV), leading to the low RSQ. However, as alcohol is increased HRV is always seen to significantly reduce.
For now, I will leave this post as is, please let me know in the comments, or message me directly if there is anything I should cover or statistical analysis you would like to see using my data. In the future I will cover some specific metrics, please consider this post an introduction into the correlation and data tracking that I do.
Thanks, Andrew