Dataset

This dataset was created by Kamal Das

Contents

BackgroundAccelerometers are widely adopted for physical activity (PA) measurement. Accelerometry data require pre-processing before entering formal statistical analyses. Many pre-processing criteria may influence PA outcomes and the processed sample, impacting results in subsequent statistical analyses.AimTo study the implications of pre-processing criteria for accelerometer data on outputs of interest in physical activity studies.MethodsWe used the ActiGraph hip-worn accelerometry data from 538 adult Latino participants. We studied four most important domains of pre-processing criteria (wear-time, minimum wear-time, intensity level, and modified bouts). We examined the true sample size in pre-processed data, the moderate-to-vigorous physical activity (MVPA) outcome, and regression coefficients of age and gender predicting MVPA.ResultsMany pre-processing criteria have minimum impact to the output of interest. However, requirements for minimum wear-time can have high influence on subsequent analyses for MVPA. High requirements for wear-time (e.g., minimum of 5 days with more than 12 hours of wear-time per day) lead to weakened statistical efficiency in estimating the relationship between potential predictors and the MVPA outcome. Intensity levels using vector magnitude triaxial counts yielded drastically different results than those using conventional vertical axis counts.ConclusionModerate changes in minimum wear-time can yield notably different output data and subsequently influence analyses assessing the impacts of interventions on MVPA behaviors. Processed data using vector magnitude and conventional vertical axis counts are not directly comparable. Sensitivity analyses using alternative pre-processing scenarios are highly recommended to verify the robustness of analyses for accelerometry data.

Features of MRI scanning and prediction model construction.