Individual Information-Centered Approach for Handling Physical Activity Missing Data

The purpose of this study was to validate individual information (II)-centered methods for handling missing data, using data samples of 118. We used a semisimulation approach to create six data sets: three physical activity outcome measurements (i.e., step counts, activity counts, and minutes of moderate to vigorous physical activity) for both groups (i:e., middle-aged adults and older adults). After analyzing each data set separately, we replaced missing values with two II-centered and two group information (GI)-centered methods. Root mean square difference (RMSD), mean signed difference, paired t tests, and Pearson correlations were used to determine the effectiveness of the various recovery methods. Overall, the II-centered methods showed smaller RMSDs than the GI-centered methods for each data set in both groups. We found no significant mean differences between the known values and the replacement values in all conditions. The II-centered methods produced better results than GI-centered methods. We determined substituting missing data points using the average of days remaining to be an accurate missing data recovery method for middle-aged adults' and older adults'pedometer and accelerometer data.