In recent years, “step counting” has become a constant pre-occupation for some people as a way of trying to tie activity to fitness goals. This has been spurred on by inexpensive digital devices that provide a suite of information on how and how much you move during the day. Many of the consumer interfaces to these data provide summary information that often focuses on daily totals and visual displays of patterns of activity if the user dives a little deeper. This suggests that when you move throughout the day is just as much of a story as how much you move in total. The purpose of this research is to explore cluster analysis of daily step count profiles using a novel dissimilarity measure that allows days that are similar but are slightly offset in timing to be grouped together. We illustrate the method with self-recorded step count activity data from a Fitbit activity tracker band over a 70 day span. The results show days that may be similar in total step counts that are separated in this analysis by looking more closely at the daily count profiles. This provides an alternative way to understand patterns in daily activity profiles that goes beyond just comparing daily totals.
Meyer, Elijah S.; Tran, Tan; and Greenwood, Mark
"Statistical methods for detecting groups of patterns in daily step count activity profiles,"
Skyline - The Big Sky Undergraduate Journal: Vol. 4
, Article 6.
Available at: http://skyline.bigskyconf.com/journal/vol4/iss1/6