Description

Six models were developed for predicting energy expenditure in children using either a right hip or left wrist-worn ActiGraph accelerometer or a combination of both monitors. Models were developed for raw acceleration data and count data separately. Models were trained using children playing active video games and walking/running on a treadmill and cross-validated in an independent sample of 15 participants who played active video games.

The following models are available. Root Mean Square Error (RMSE) is also provided.

• Artificial neural network for hip-worn monitor using count data (1.2 METs)

• Artificial neural network for hip-worn monitor using raw data (1.9 METs)

• Artificial neural network for wrist-worn monitor using count data (~2.2 METs*)

• Artificial neural network for wrist-worn monitor using raw data (3.4 METs)

• Artificial neural network for hip- and wrist-worn monitors (combined) using count data (~1.8 METs*)

• Artificial neural network for hip- and wrist-worn monitors (combined) using raw data (1.7 METs)

*estimated from Figure

Development/Validation

Sample: 27 (15 males) healthy children with a mean age 11.6±1.0 years

Setting: Laboratory

Activities: Active video games, treadmill walking and running

Criterion: MetaMax 3b (VO₂) in development or OxyCon Mobile (VO₂) in cross-validation

Accelerometer(s): ActiGraph GT3X+ at the right hip and left wrist

Validation approach: Cross-validation in an independent sample of 34 children

Phase Designation

(What's this?)

This model is in Phase 0, 1, and 4.

Instructions

The R code can be used to read in a csv file with the necessary features and then use the chosen model to predict energy expenditure in METs for 15-sec windows. There is a separate file for each of the six models- the hip count, hip raw, wrist count, wrist raw, combination count, and combination raw models. There is an example data sheet with the variables described in the data dictionary that need to be present in the csv file (this data sheet includes all wear locations and count/raw data). 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 data

Data dictionary

Hip count model

Hip raw model

Wrist count model

Wrist raw model

Combination count model

Combination raw model

Note: Compared to the original source, code was edited for consistency, data dictionary added, models and example data were unchanged

Source Information

Reference:

Montoye, A. H., Clevenger, K. A., Mackintosh, K. A., McNarry, M. A., & Pfeiffer, K. A. (2019). Cross-validation and comparison of energy expenditure prediction models using count-based and raw accelerometer data in youth. Journal for the Measurement of Physical Behaviour, 2(4), 237-246. https://doi.org/10.1123/jmpb.2018-0011 Link to paper

Original source for code:

https://drive.google.com/file/d/1SlnXJBh6WUpxJJAjAovVbNw8hW54PhbZ/view

Corresponding author: Alexander H.K. Montoye, montoyeah@alma.edu

Contact

Kimberly Clevenger at accelerometerrepository@gmail.com