A refined 2-regression model for the ActiGraph accelerometer
Crouter, 2010
Description
An updated 2-regression method is presented in which the coefficient of variation in counts/10-s from a hip-worn ActiGraph is used to determine which of two equations is used to predict METs. This approach was developed in adults performing structured activities over a range of intensities. Mean bias was 0.1 METs and average Root Mean Square Error (RMSE) across activities was 0.84 METs.
Development/Validation
Sample: 48 (24 males) healthy adults, 21-69 years of age
Setting: Laboratory, university, or home
Activities: Activities of daily living, overground walking and running, sports, stairs, stationary
Criterion: Cosmed K4b2 (VO2)
Accelerometer(s): ActiGraph 7164 on right hip
Validation approach: Holdout of 15 participants
Instructions
The models can be implemented as described below or using the TwoRegression R package.
The coefficient of variation (CV) for each 10-second epoch and all possible combinations of the surrounding five 10-sec epochs is calculated. So, the CV is calculated for the 10-sec epoch of interest and the 1) five 10-sec epochs before, 2) four 10-sec epochs before and one 10-sec epoch after, 3) three 10-sec epochs before and two 10-sec epochs after, 4) two 10-sec epochs before and three 10-sec epochs after, 5) one 10-sec epoch before and four 10-sec epochs after, and 6) five 10-sec epochs that followed. Then, the lowest CV for each 10-second epoch is used to determine the appropriate equation for predicting METs/min.
Source Information
Crouter, S. E., Kuffel, E., Haas, J. D., Frongillo, E. A., & Bassett Jr, D. R. (2010). A refined 2-regression model for the ActiGraph accelerometer. Medicine and Science in Sports and Exercise, 42(5), 1029. https://doi.org/10.1249/MSS.0b013e3181c37458 Link to Paper
Corresponding author: Scott Crouter, Scott.crouter@umb.edu
Contact
Kimberly Clevenger at accelerometerrepository@gmail.com