mMARCH.AC - Processing of Accelerometry Data with 'GGIR' in mMARCH
Mobile Motor Activity Research Consortium for Health
(mMARCH) is a collaborative network of studies of clinical and
community samples that employ common clinical, biological, and
digital mobile measures across involved studies. One of the
main scientific goals of mMARCH sites is developing a better
understanding of the inter-relationships between
accelerometry-measured physical activity (PA), sleep (SL), and
circadian rhythmicity (CR) and mental and physical health in
children, adolescents, and adults. Currently, there is no
consensus on a standard procedure for a data processing
pipeline of raw accelerometry data, and few open-source tools
to facilitate their development. The R package 'GGIR' is the
most prominent open-source software package that offers great
functionality and tremendous user flexibility to process raw
accelerometry data. However, even with 'GGIR', processing done
in a harmonized and reproducible fashion requires a non-trivial
amount of expertise combined with a careful implementation. In
addition, novel accelerometry-derived features of PA/SL/CR
capturing multiscale, time-series, functional, distributional
and other complimentary aspects of accelerometry data being
constantly proposed and become available via non-GGIR R
implementations. To address these issues, mMARCH developed a
streamlined harmonized and reproducible pipeline for loading
and cleaning raw accelerometry data, extracting features
available through 'GGIR' as well as through non-GGIR R
packages, implementing several data and feature quality checks,
merging all features of PA/SL/CR together, and performing
multiple analyses including Joint Individual Variation
Explained (JIVE), an unsupervised machine learning dimension
reduction technique that identifies latent factors capturing
joint across and individual to each of three domains of
PA/SL/CR. In detail, the pipeline generates all necessary
R/Rmd/shell files for data processing after running 'GGIR'
(v2.4.0) for accelerometer data. In module 1, all csv files in
the 'GGIR' output directory were read, transformed and then
merged. In module 2, the 'GGIR' output files were checked and
summarized in one excel sheet. In module 3, the merged data was
cleaned according to the number of valid hours on each night
and the number of valid days for each subject. In module 4, the
cleaned activity data was imputed by the average Euclidean norm
minus one (ENMO) over all the valid days for each subject.
Finally, a comprehensive report of data processing was created
using Rmarkdown, and the report includes few exploratory plots
and multiple commonly used features extracted from minute level
actigraphy data. Reference: Guo W, Leroux A, Shou S, Cui L,
Kang S, Strippoli MP, Preisig M, Zipunnikov V, Merikangas K
(2022) Processing of accelerometry data with GGIR in Motor
Activity Research Consortium for Health (mMARCH) Journal for
the Measurement of Physical Behaviour, 6(1): 37-44.
