Exercise and Body Weight During Post-Chemotherapy Recovery: An N-of-1 Study

Self-directed n-of-1 statistical study, built from 2.5 years of daily self-tracked data

• Applied Statistics

• Pandas
• NumPy
• SciPy
• Matplotlib

Overview

Over two and a half years of recovery from Stage 4 brain cancer and the chemotherapy-induced nerve damage that came with it, I logged my body weight, exercise, and diet every day. This project turns that log into a single-subject (n-of-1) statistical study asking a simple question: once calorific intake is controlled for, does exercising actually change how much my weight moves day to day?

Building the Dataset

Daily body weight readings came from an Omron BF508 digital scale, exercise sessions were logged via Strava (recorded on an Apple Watch Series 5 or Garmin Vívoactive 6), and diet was tracked in MyFitnessPal. I merged the three sources by date, labelled each day as an exercise or non-exercise day (discarding sessions that burned under 150 calories as estimated by Strava), and restricted the dataset to days with ≤100g of carbohydrate intake to hold diet composition roughly constant. After removing incomplete rows and outliers in daily weight change, I still had an imbalance in calorific intake between the two groups. To fix that, I stratified days into 10-calorie bins and drew matched samples from each bin. This left 41 observations per group for exercise and non-exercise days.

Testing for Difference

A Shapiro-Wilk test showed the exercise group's daily weight changes were normally distributed while the non-exercise group's weren't, so I used the Mann-Whitney U test as the primary comparison rather than a t-test. It returned U = 874.5, p = 0.756, failing to reject the null hypothesis that exercise has no effect on daily weight change.