Using arctools package to compute physical activity
summaries
The arctools functions process one file with
accelerometry data at a time.
Reading PA data
Four CSV data sets with minute-level activity counts data are
attached to the arctools package. The data file names are
stored in extdata_fnames object that becomes available once
the arctools package is loaded.
Below, we defined fpath to be a path to one of the
minute-level activity counts data files. fread() reads
minute-level activity counts data file while conveniently skipping first
few rows with meta data, and then as.data.frame() converts
the read data into a data frame object. The read-in data is assigned to
dat variable. head() and tail()
get first few and last few rows of dat, respectively.
library(arctools)
library(data.table)
library(dplyr)
library(ggplot2)
library(lubridate)
## Read one of the data sets
fpath <- system.file("extdata", extdata_fnames[1], package = "arctools")
dat <- as.data.frame(fread(fpath))
rbind(head(dat, 3), tail(dat, 3))
#> Axis1 Axis2 Axis3 vectormagnitude timestamp
#> 1 1021 1353 2170 2754 2018-07-13 10:00:00
#> 2 1656 1190 2212 3009 2018-07-13 10:01:00
#> 3 2540 1461 1957 3524 2018-07-13 10:02:00
#> 10078 0 0 0 0 2018-07-20 09:57:00
#> 10079 0 0 0 0 2018-07-20 09:58:00
#> 10080 0 0 0 0 2018-07-20 09:59:00The data columns are:
Axis1- sensor’s X axis minute-level counts data,Axis2- sensor’s Y axis minute-level counts data,Axis3- sensor’s Z axis minute-level counts data,vectormagnitude- minute-level counts data defined assqrt(Axis1^2 + Axis2^2 + Axis3^2),timestamp- time-stamps corresponding to minute-level measures.
## Plot activity counts
## Format timestamp data column from character to POSIXct object
ggplot(dat, aes(x = ymd_hms(timestamp), y = vectormagnitude)) +
geom_line(size = 0.3, alpha = 0.8) +
labs(x = "Time", y = "Activity counts") +
theme_gray(base_size = 10) +
scale_x_datetime(date_breaks = "1 day", date_labels = "%b %d")
Computing summaries with activity_stats method
acc <- dat$vectormagnitude
acc_ts <- ymd_hms(dat$timestamp)
activity_stats(acc, acc_ts)
#> n_days n_valid_days wear_time_on_valid_days tac tlac ltac
#> 1 8 4 1440 2826648 6429.838 14.8546
#> astp satp time_spent_active time_spent_nonactive
#> 1 0.1781782 0.09516215 499.5 940.5
#> no_of_active_bouts no_of_nonactive_bouts mean_active_bout mean_nonactive_bout
#> 1 89 89.5 5.61236 10.50838Output explained
To explain activity_stats method output, we first define
the terms activity count, active/non-active minute,
active/non-active bout, and valid day.
- Activity count (AC) - a minute-level metric of PA volume.
- Active minute - a minute with AC equal or above a fixed threshold;
for wrist-worn Actigraph
we use AC>=1853 (method’s default). - Non-active (sedentary) minute - a minute with AC below a fixed
threshold; for wrist-worn Actigraph
we use AC<1853 (method’s default). - Active bout - a sequence of 1 or more consecutive active minute(s).
- Non-active bout - a sequence of 1 or more consecutive non-active minute(s).
- Valid day - a day with no more than 10% of the non-wear time (see
Details in
?activity_stats).
Meta information:
n_days- number of days (unique day dates) of data collection.n_valid_days- number of days (unique day dates) of data collection determined as valid days.wear_time_on_valid_days- average number of wear-time minutes across valid days.
Summaries of PA volumes metrics:
tac- TAC, Total activity counts per day - sum of AC measured on valid days divided by the number of valid days.tlac- TLAC, Total-log activity counts per day - sum of log(1+AC) measured on valid days divided by the number of valid days. Here ‘log’ denotes the natural logarithm.ltac- LTAC, Log-total activity counts - natural logarithm of TAC.time_spent_active- Average number of active minutes per valid day.time_spent_nonactive- Average number of sedentary minutes per valid day.
Summaries of PA fragmentation metrics:
astp- ASTP, active to sedentary transition probability on valid days.satp- SATP, sedentary to active transition probability on valid days.no_of_active_bouts- Average number of active minutes per valid day.no_of_nonactive_bouts- Average number of sedentary minutes per valid day.mean_active_bout- Average duration (in minutes) of an active bout on valid days.mean_nonactive_bout- Average duration (in minutes) of a sedentary bout on valid days.