How Many Missing Days Can Be Recovered? Children's Step-Count Data

Friday, April 26, 2013
Exhibit Hall Poster Area 2 (Convention Center)
Minsoo Kang and Youngdeok Kim, Middle Tennessee State University, Murfreesboro, TN

Background/Purpose The effectiveness of individual information-centered (IIC) approach for recovering the missing step-count data has been well established. However, a question still remains of how many days of replacement can be made. The purpose of this study was to examine the threshold of the number of missing days replaced for valid estimates of step-count data for children.

Method A total of 333 developing children, aged 2-15 years old, had 7 consecutive days of complete step-count data using a StepWatch Activity Monitor. Missing datasets (1-day through 5-day missing) were created 10,000 times in which all missing values were replaced using the IIC approach. A 7-day average was calculated for each dataset, including the complete dataset. One-way repeated measures ANOVA was used to determine the differences between 1-day through 5-day missing datasets and the complete dataset. Mean Absolute Percent Error (MAPE) and 95% CI were also computed.

Analysis/Results Mean (SD) daily step counts for the complete 7-day dataset was 7754 (2735). Mean (SD) values for the 1-day through 5-day missing datasets were 7723 (2750), 7774 (2752), 7836 (2901), 7754 (2930), and 7616 (2953), respectively (p>.05). The lower MAPEs were estimated for 1-day missing (5.2%, 95% CI 4.4–6.0), 2-day missing (8.4%, 95% CI 7.0–9.8), and 3-day missing (7.8%, 95% CI 7.2–8.5) while all others were greater than 10%.

Conclusions Based on the MAPE results, the IIC approach can be effectively utilized for recovering children's pedometer missing data up to maximum of three missing days out of seven days (MAPE<10%).