Relatively little attention has been paid to the effects of nonparticipation on data quality in population-based studies that use accelerometers to measure physical activity. We examine these issues using data from the 2013 Longitudinal Internet Studies for the Social Sciences (LISS) panel and 2013–2014 National Health and Nutrition Examination Survey (NHANES) accelerometer studies, both of which collected survey data in advance and therefore permit comparisons of self-reported physical activity between participants and nonparticipants to the accelerometer studies. While individuals with high levels of self-reported physical activity are overrepresented in the participant samples, the differences are modest in both studies. However, in the LISS panel this difference led to overestimates of physical activity that are not fully corrected by propensity score weighting adjustments (i.e., non-ignorable selection bias). This finding underscores the importance of assessing the potential influence of nonparticipation on accelerometer-derived estimates of physical activity.

中文翻译：

---

基于加速度计的研究中的不参与偏差和倾向评分的使用


在使用加速度计测量身体活动的基于人群的研究中，相对较少关注不参与对数据质量的影响。我们使用 2013 年社会科学纵向互联网研究 (LISS) 小组和 2013-2014 年国家健康与营养检查调查 (NHANES) 加速计研究的数据来研究这些问题，这两项研究都提前收集了调查数据，因此可以对自我进行比较-报告加速度计研究参与者和非参与者之间的身体活动。虽然自我报告身体活动水平高的个体在参与者样本中所占比例过高，但两项研究中的差异并不大。然而，在 LISS 面板中，这种差异导致了对体力活动的高估，而这种高估并未通过倾向得分权重调整（即不可忽略的选择偏差）得到完全纠正。这一发现强调了评估不参与对加速度计得出的身体活动估计的潜在影响的重要性。