Measuring the level of physical activity (PA) of the general population accurately is important to guide public health policies. Instead of relying on self-reported measures which are prone to social desirability and recall biases, studies increasingly provide accelerometers to study participants for passively measuring PA. After the raw accelerometer data are collected, researchers have to go through a series of processing steps, such as removing extreme values and aggregating the data into fixed time intervals. A concern in accelerometer studies is identifying non-adherence, i.e., participants taking off the device, resulting in non-wear time. The accurate processing and calculation of non-wear time in the accelerometer data is important to ensure data quality. Although prior research has identified how a subset of processing decisions affect the measurement of non-wear time, there is currently only limited research about how decisions across the entire processing pipeline can affect the estimation of non-wear time in tri-axial accelerometer studies of the general population.
