In my life with diabetes, I have found that meaningful shifts in metabolism do not happen day-to-day or week-to-week, but in the range of months and years. This experiment takes a look at average daily metabolism in a calendar year.
My glucose meter data dates back to March 1st, 2012 during my first year at Turner Broadcasting. This time in my life saw the start of a more stable daily routine, and with it, a drastic change in my diet, exercise, and resulting glucose levels.
Getting started, I plotted every day in a year stacked one on top of the other. This visual was compelling, and my eye immediately started to tease out patterns. The human eye seeks structure in chaos, and while I coule tell that there were discernable patterns, I needed some further design to tease them out.
In order to guide the eye towards a more accurate trendline, I decided to take a look at median values. For every hour in every day across a year, I've calculated the median (mean-quartile) reading from this set. This way, the average is not computed, but an actual representative value. For each of these median values, I computed the median time of capture for the hour block as well, increasing the accuracy of the line shape.
For early morning hours with only a few readings, I padded the values in each hour block so that there were at least 60 readings per hour set. The 'padded' values are the average of all readings for the year. Since these waking readings are almost 100% outlier readings (read: I felt weird, woke up and checked my levels), this helped to damp the values so that the line did not skew dramatically high or low based on such limited data.
For the final step, I highlighted the lowest and highest values to give a sense of context for the line.