穿戴式裝置提供了收集人類身體活動信息的機會。然而受試者的意願和其他潛在行為,將會使參數估計時產生不可忽略的偏差。在此類型資料的分析中,研究人員通常使用半母數或無母數方法來避免模型錯誤所造成的偏差。但另一方面,有母數方法可以通過模型選擇來控制這種偏差,並且可以大幅的提升運算效率。在本文中,我們提供模擬研究來比較半母數方法和有母數方法的表現,並將我們的方法應用於來自美國國家健康和營養檢查調查的穿戴式裝置數據。;Wearable devices provide the opportunity to collect information of human being′s physical activity. However, there is non-negligible deviation from the subject′s willingness and other potential behaviors. In wearable device data analysis, researchers usually utilize semi-parametric or nonparametric approaches to avoid the bias from model misspeci cation. On the other hand, parametric approaches can control such bias by model selection, and can reduce computing time signi cantly. In this paper, we provide simulation studies to compare the performance of the semiparametic and parametric approaches. We apply our approach to the wearable device data from National Health and Nutrition Examination Survey is USA.