Description

Four “Sojuourn” models were developed for estimating energy expenditure in children using either activity counts (vector magnitude) or Euclidean Norm Minus One (ENMO) from an ActiGraph worn on the right hip or non-dominant wrist. Models, which are a combination of decision tree and artificial neural network, were developed in the laboratory and cross-validated using an independent, free-living sample.

The following four models are available. Percent accuracy is also provided for both the original laboratory-based development phase and the free-living cross-validation phase.

• Hip count (development: 69.7%, cross-validation: 56.7%)

• Hip raw (development: 71.6%, cross-validation: 49.3%)

• Wrist count (development: 68.5%, cross-validation: 50.7%)

• Wrist raw (development: 70.3%,cross-validation: 49.4%)

Development/Validation

Sample: 54 (21 males) healthy children, 7-13 y of age

Setting: Laboratory

Activities: Not reported

Criterion: Oxycon Mobile (VO₂)

Accelerometer(s): ActiGraph GT3X+ on right hip and non-dominant wrist for model development, but ActiGraph GT3X-BT on right hip and both wrists for cross-validation

Validation approach: Independent sample of 27 children in free-living

Phase Designation

(What's this?)

This model is in Phase 4.

Instructions

This model can be implemented using the Sojourn R package.

The R code can be used to read in a csv data file, then use the Sojourn model to predict energy expenditure for each 1-s window. There are four example data sheets: count and raw versions in the format needed for the Sojourn model which requires you to have already added some variables and count and raw versions in the format exported by ActiGraph which can be used with the R code to add all required variables. Further instructions are in the R code. More information about R is found here.

This code relies on R software which can be downloaded for free at https://www.r-project.org/

Attached Files

Download all files as a .zip or download individual files below.

Code

Example formatted count data sheet

Example formatted raw data sheet

Unprocessed count data

Unprocessed raw data

Source Information

Reference:

Hibbing, P. R., Ellingson, L. D., Dixon, P. M., & Welk, G. J. (2018). Adapted Sojourn Models to Estimate Activity Intensity in Youth: A Suite of Tools. Medicine and science in sports and exercise, 50(4), 846-854. https://www.doi.org/10.1249/MSS.0000000000001486 Link to paper

Original source for code: https://cran.r-project.org/web/packages/Sojourn/Sojourn.pdf

Corresponding author: Paul Hibbing, phibbing@vols.utk.edu

Repository contact: Kimberly A. Clevenger, clevengerkimberly@gmail.com

Contact

Kimberly Clevenger at accelerometerrepository@gmail.com