高齢者の週の交通行動を予測するためには何日間の調査が必要か？

抄録

　This study determined how many days of data were needed to predict weekly travel behavior in older adults. Seven days of travel behavior data were collected from 142 older adults aged ≥ 65 years, who participated in health promotion classes conducted in Shizuoka, Japan. Using these data, we calculated the time spent walking, cycling, and using cars and public transportation each day. We then assessed intraclass correlations among the combinations of days with regard to travel behavior and conducted multiple regression analysis to examine the extent to which these data explained weekly travel behavior based on the estimated adjusted coefficients of determination. To ensure values ≥ 0.80 of the intra-class correlations, information on car use and other travel behaviors (walking, cycling, and public transportation use) was needed for at least four and three days, respectively. The range of adjusted R2 was 0.77–0.89 for walking, 0.76–0.94 for cycling, and 0.74–0.93 for public transportation use based on travel data for three days, and 0.76–0.93 for car use based on data for four days. Our results suggest that the specific single-day or twoday data collection that Japanese travel surveys commonly administered may be not sufficient to predict the weekly travel behavior of older adults, whereas information about travel behaviors over three-four days would greatly enhance the predictability of such behavior with an acceptable level of accuracy.