Detecting Changes in Daily Step Count

Due to the widespread adoption of wearable devices over the past decade, individuals now have unprecedented access to high-frequency personal health data. A 2023 NIH survey revealed that “one in three Americans uses a wearable device, such as a smart watch or band, to track their health and fitness.” The data collected from these devices can be used to detect changes in a person’s health or lifestyle. I don’t personally own any wearable devices, but for nearly a decade I have used my phone to record the steps I take in a given day. This data is bundled with the mich package as the daily_steps data set. Below I show how I used mich to detect the changes in my life that manifest in daily_steps. In Berlind, Cappello, and Madrid Padilla (2025), we showed how to extend the MICH model to count data, but seeing as this feature is not yet implemented in the mich package, I instead model the change-points as joint changes to the mean and variance of daily_steps$total_steps and set J_auto == TRUE to automatically select the number of mean-variance components to include in the model.

library(mich)
daily_steps$day <- as.Date(daily_steps$day)

step_fit <- mich(daily_steps$total_steps, J_auto = TRUE, tol = 1e-12,
                 max_iter = Inf, restart = FALSE)

After fitting mich() to the step data, we see that there are 12 changes detected, each of which is annotated in the table below:

Estimated Change	90% CS Size	Annotation
May 24, 2017	47	I lived in NYC from 2016 to 2018 and walked much more during the summer. This change-point appears to be this seasonal change (note the wide 90% credible set that extends from March to June).
April 06, 2018	34	Again, this change seems to be the switch from winter to summer (note that MICH apparently misses the seasonal decrease in the winter of 2017).
August 21, 2018	12	I moved from NYC to Durham, NC at the beginning of August 2018, to begin a masters degree program at Duke (orientation was August 21, 2018).
August 21, 2018	12	I moved from NYC to Durham, NC at the beginning of August 2018, to begin a masters degree program at Duke (orientation was August 21, 2018).
August 13, 2019	1	From 8-12-2019 to 8-24-2019 I traveled to Ireland and Scotland with my now wife.
August 24, 2019	1	Return from Ireland/Scotland.
March 17, 2020	4	On 3-17-2020 I flew to Miami to spend the first phase of lock-down with my wife and her family.
June 12, 2020	2	On 6-11-2020 I moved to Boston.
September 05, 2020	5	On 9-2-2020 I began a PhD in economics at BU.
May 10, 2021	7	On 5-8-2021 I took my last exam at BU and left the program to transfer to the Department of Statistics at UCLA.
June 30, 2023	3	From 6-29-2023 to 7-13-2023 I traveled to Prague, Vienna, and Greece for a friend’s wedding.
July 10, 2023	5	Returned from Greece.
May 09, 2025	27	This change captures the fact that in May I hunkered down and barely left my desk in order to get the preprint of my paper up on arXiv 💀.

Davis Berlind

2025-09-04