The project uses raw accelerometry data from ~700 participants in Baltimore Longitudinal Study of Aging (BLSA), each monitored for aporox. 1 week with a wrist-worn ActiGraph sensor.
The goal of the project is to:
Develop and use R software to aggregate raw data at minute-level with open-source measures: MIMS, ENMO, MAD, AI.
Quantify association between AC and open-source measures marginally and conditionally on age, sex and BMI.
Harmonize minute-level AC with open-source measures via one-to-one mapping.
Reproduce some of the published BLSA results that used AC with the use of the open-source measures.
For practitioners, potentially most useful R code scripts and result files are referenced below.
R code script to generate raw data quality check flags: code/data_preprocessing/mat_to_minute_quality_flag.R
R code script to compute MIMS, MAD, AI: code/data_preprocessing/mat_to_open_source_measures.R
R code script to compute ENMO (with data calibration step): code/data_preprocessing/mat_calibrated_to_open_source_measures.R
R code script to compute valid minute and valid day flags and to filter the participants: code/data_preprocessing/prepare_measures_masterfile.R
R code script to perform data imputation: code/data_preprocessing/prepare_measures_masterfile_winsorized_imp.R
CSV table with model-fitted values of MIMS, ENMO, MAD, AI for a range of AC values: results_public/mapping_between_measures_FITTED.txt
R code script with fast mapping functions between AC and MIMS, ENMO, MAD, AI: code/data_preprocessing/measures_mapping_FUNC.R