Effect Of Different Epoch Lengths On Activity Counts Using The Actigraph GT1M Accelerometer
Minsoo Kang, Tiago V. Barreira, Elizabeth A. Holbrook, David A. Rowe · Medicine & Science in Sports & Exercise · 2010
PURPOSE: To determine if Actigraph activity counts collected in 1-sec and 15-sec epochs then integrated to 1-min epochs provide the same measurements as an epoch length set to 1-min. METHODS: Six adults participated in this study and completed three walking trials wearing four Actigraph accelerometers. Two accelerometers were set at 1-min, one was set at 1-sec, and one was set at 15-sec epochs. One accelerometer set at 1-min epoch was worn on the left side of the hip and another on the right. The 1-sec and 15-sec accelerometers were worn on opposite sides of the hip, alternating positions by participants. Participants walked at a moderate intensity for approximately 1 mile on an indoor track, with a 2-min rest at the halfway point. A total of 153 minutes of data were collected. The 1-sec and 15-sec data were then re-integrated to 1-min using the ActiLife software. Paired t-test was used to determine if there was a significant difference between the 1-min accelerometers on the left and right sides. Because there was no significant difference (t(152) = 0.99, p =.32), the data were combined and the further analyses were done independent of which side the accelerometer was worn. Absolute percentage error (APE), Pearson correlation, and paired t-test were computed between the 1-sec and 1-min and between 15-sec and 1-min epochs. RESULTS: The average activity counts for the re-integrated 1-sec and the 1-min epochs were 4,060 ± 1,239 and 3,972 ± 1,121, respectively. For the re-integrated 15-sec and the 1-min epochs, they were 4,113 ± 1,131 and 4,041 ± 1,069, respectively. APE was 8.1 ± 6.0% for 1-sec and 8.9 ± 5.4% for the 15-sec epochs, when compared to 1-min epochs. Pearson correlations were high,.94 and.93 between 1-sec and 1-min and between 15-sec and 1-min epochs, respectively. Paired t-test showed significant differences in activity counts between 1-sec and 1-min epochs, t(152) = 2.6, p =.01, and between 15-sec and 1-min epochs, t(152) = 2.2, p =.03. Both mean differences corresponded to small effect sizes (Cohens's d = 0.07 for both comparisons). CONCLUSIONS: Future researchers should be cautious in setting short epoch lengths when collecting accelerometer data, if the intention is to later re-integrate the data into 1-min epochs.