Accuracy of a difference in 24-hour physical behaviors between two groups: effects of balanced and unbalanced group sizes across scenarios with varying numbers of measurement days
Year: 2026
Type: article
Abstract: Background Too little data can lead to uncertain and potentially biased group differences in physical behaviors, while excess data may waste resources. In order to examine the trade-off between measurement efforts and study performance, we examined the accuracy (the combined effect of trueness and precision) of the difference in 24-hour physical behaviors between overweight and normal-weight office workers of the number of participants and the number of measurement days per participant in balanced and unbalanced data sets. Methods Time spent sitting, standing, moving, and sleeping was monitored for 7 days using a thigh-worn accelerometer in 103 overweight (BMI ≥ 25 kg/m 2 ) and 122 normal-weight (< 25 kg/m 2 ) office workers. Using this parent dataset, we simulated, using bootstrap resampling, data collection strategies with 96 (8 × 4 × 3) combinations of the total number of participants (18, 36, 54, 72, 90, 108, 126, 144), working days per participant (1, 2, 3, 4), and balancedness (ratios 1:1, 1:2, and 2:1 between overweight and normal-weight participants). For each sampling strategy, trueness (the size) and precision (the 95% confidence interval width) of the group difference, both expressed in compositional space, was compared to the parent dataset, and the likelihood of obtaining statistically significant group differences was assessed as a measure of the combined effect of trueness and precision. Results Increasing the number of participants per group did not change the estimated behavior difference at a specific number of working days. Measuring only one day introduced bias to the difference, while more days produced estimates closer to the observed “true” difference in the parent dataset. Precision improved with more participants and days, and a balanced study generally yielded more precise estimates than an unbalanced study with the same total number of measurements. The likelihood of arriving at a significant group difference was influenced by the number of both participants and days. Conclusions Trueness (i.e., bias) was influenced by the number of days, but not the number of participants. In contrast, precision was influenced by the number of participants, days, and balancedness of the study population. Thus, it is important to consider the trade-offs between the number of participants, days and balancedness in future studies, and even to consider the costs of collecting data from more participants or more days per participant in either group. Trial registration Not applicable.
Language: English
Topic: Physical Activity and Health
Field: Medicine
Domain: Health Sciences
Open Access status: hybrid
Estimated APC paid: $1,890
Funder: Högskolan i Gävle
OpenAlex ID: W7232399266
