Type: Package
Title: Develop and Apply Two-Regression Algorithms
Version: 1.1.1
Depends: R (≥ 3.5.0)
Description: Facilitates development and application of two-regression algorithms for research-grade wearable devices. It provides an easy way for users to access previously-developed algorithms, and also to develop their own. Initial motivation came from Hibbing PR, LaMunion SR, Kaplan AS, & Crouter SE (2018) <doi:10.1249/MSS.0000000000001532>. However, other algorithms are now supported. Please see the associated references in the package documentation for full details of the algorithms that are supported.
License: GPL-3
Encoding: UTF-8
LazyData: true
Imports: dplyr (≥ 0.5.0), ggplot2 (≥ 3.4.0), gridExtra (≥ 2.3), lubridate, magrittr (≥ 1.5), PAutilities (≥ 1.1.0), pROC (≥ 1.16.0), RcppRoll, rlang, stats, tidyr
RoxygenNote: 7.1.2
Suggests: knitr, rmarkdown, testthat (≥ 3.0.0)
VignetteBuilder: knitr
URL: https://github.com/paulhibbing/TwoRegression
BugReports: https://github.com/paulhibbing/TwoRegression/issues
Config/testthat/edition: 3
NeedsCompilation: no
Packaged: 2025-11-15 12:38:45 UTC; paulh
Author: Paul R. Hibbing [aut, cre], Vincent T. van Hees [ctb]
Maintainer: Paul R. Hibbing <paulhibbing@gmail.com>
Repository: CRAN
Date/Publication: 2025-11-19 19:50:13 UTC

Perform leave-one-participant-out-cross-validation on a two-regression algorithm

Description

Perform leave-one-participant-out-cross-validation on a two-regression algorithm

Usage

DualCP_LOSO(
  subject_var = "id",
  data,
  model,
  MET_var = "MET_RMR",
  activity_var = "Behavior",
  verbose = FALSE,
  trace = FALSE
)

fold(x, subject_var, data, model, MET_var, activity_var, trace)

get_cv_predictions(model, fold_data, cv_data)

get_fold_model(formula_string, fold_data, level = c("walkrun", "intermittent"))

get_classifications(
  data,
  model,
  numeric = TRUE,
  labels = c("SB", "walkrun", "intermittent")
)

Arguments

subject_var

character. Variable name that distinguishes between participants

data

the full data set to cross-validate

model

a TwoRegression object formed with fit_2rm on which to perform the cross-validation

MET_var

character. The outcome variable name (in metabolic equivalents)

activity_var

character. The variable name for the activity being performed

verbose

logical. Print updates?

trace

logical. Print information about each iteration?

x

character. The id to hold out

fold_data

the validation data set

cv_data

the holdout (i.e., cross-validation) data set

formula_string

character. Formula to apply in call to lm

level

character. Classification subset to include in call to lm

Value

A data frame with predictions obtained from leave-one-participant-out-cross-validation

Note

This function will not work for TwoRegression objects formed from previously-published research. The TwoRegression object needs to have more information than is available in those cases in order to perform cross-validation, and this is sensible, since there is no reason or way to re-perform cross-validation on an already-finalized algorithm.

Apply an existing two-regression model

Description

Acts as a generic function that dispatches to specific implementation functions (crouter_2006, crouter_2010, crouter_2012, or hibbing_2018)

Usage

TwoRegression(
  AG,
  method = c("Crouter 2006", "Crouter 2010", "Crouter 2012", "Hibbing 2018"),
  verbose = FALSE,
  ...
)

crouter_2006(AG, movement_var, time_var, ...)

crouter_2010(AG, movement_var, time_var, ...)

crouter_2012(AG, movement_var, time_var, model, check = TRUE, ...)

hibbing_2018(
  AG,
  accel_var = "ENMO",
  gyro_var = "GVM",
  direction_var = "Direction",
  time_var = "Timestamp",
  site = c("Hip", "Left Wrist", "Right Wrist", "Left Ankle", "Right Ankle"),
  algorithm = 1:3,
  smooth = FALSE,
  verbose = FALSE,
  ...
)

Arguments

AG

data frame of ActiGraph data

method

character scalar telling which model to apply to the data. Currently supported selections are "Crouter 2006", "Crouter 2010", "Crouter 2012", and "Hibbing 2018". See following subsections

verbose

logical. Print updates to console?

...

arguments passed to the implementation function indicated by method, then further passed to predict.TwoRegression and (when method == "Hibbing 2018") to smooth_2rm as well

movement_var

character scalar. Name of the movement variable (default is "Axis1")

time_var

character scalar. Name of the timestamp variable (required to verify that input epoch length is 10 seconds)

model

character scalar. One of "VA" (for the vertical axis, youth-specific model of Crouter et al., 2012) or "VM" (for the vector magnitude youth-specific model from the same paper)

check

Logical. Should a warning be issued that will prompt you to check that the selected model matches the selected movement variable?

accel_var

Character scalar. Name of accelerometer variable to operate on (expected format is Euclidian Norm Minus One, in milli-gravitational units)

gyro_var

character scalar. Name of gyroscope variable to operate on (expected format is gyroscope vector magnitude, in degrees per second)

direction_var

character scalar. Name of magnetometer variable to operate on (expected format is a vector of directions, likely produced by AGread::classify_magnetometer)

site

character scalar or vector of attachment sites (more than one may be desired, e.g., if results are to be compared from running both of the wrist-specific algorithms)

algorithm

An integer/numeric scalar or vector giving the algorithm(s) to apply to the data from the primary accelerometer and (if applicable) IMU. Must be 1 (accelerometer only), 2 (accelerometer and gyroscope), 3 (accelerometer, gyroscope, and magnetometer), or any combination thereof

smooth

logical. Should data be averaged over a longer time period after processing?

Value

The original data appended with columns giving activity classification (sedentary, ambulatory, or intermittent) and energy expenditure (i.e, METs)

See Also

Crouter et al. (2006, J Appl Physiol) Crouter et al. (2010, Med Sci Sports Exerc) Crouter et al. (2012, Med Sci Sports Exerc) Hibbing et al. (2018, Med Sci Sports Exerc)

apply_two_regression_hibbing18 smooth_2rm

Examples


## Datasets

  data(count_data, package = "TwoRegression")
  data(all_data, package = "TwoRegression")

## Crouter 2006-2012 models

  TwoRegression(
    count_data, "Crouter 2006",
    movement_var = "Axis1", time_var = "time"
  )

  TwoRegression(
    count_data, "Crouter 2010",
    movement_var = "Axis1", time_var = "time"
  )

  TwoRegression(
    count_data, "Crouter 2012", movement_var = "Axis1",
    time_var = "time", model = "VA", check = FALSE
  )

  TwoRegression(
    count_data, "Crouter 2012", movement_var = "Vector.Magnitude",
    time_var = "time", model = "VM", check = FALSE
  )

## Hibbing 2018 models (can be vectorized)

  all_data$ENMO_CV10s <- NULL
  all_data$GVM_CV10s  <- NULL
  all_data$Direction  <- NULL

  result <- TwoRegression(
    all_data, "Hibbing 2018", gyro_var = "Gyroscope_VM_DegPerS",
    direction_var = "mean_magnetometer_direction",
    site = c("Left Ankle", "Right Ankle"), algorithm = 1:2
  )

  utils::head(result)

Develop and Apply Two-Regression Algorithms

Description

The TwoRegression package is designed to make working with two-regression algorithms quick, easy, and accurate.

Details

Originally, the package was designed to house the algorithms created by Hibbing et al. (2018). Since then, support has been added for other algorithms, including Crouter et al. (2006), Crouter et al. (2010), and Crouter et al. (2012). Functionality has also been added to develop and cross-validate new two-regression algorithms. The package RcppRoll has also been invoked to speed up rolling coefficient of variation calculations.

Associated References

Hibbing PR, LaMunion SR, Kaplan AS, & Crouter SE (2018). Estimating energy expenditure with ActiGraph GT9X Inertial Measurement Unit. Medicine and Science in Sports and Exercise. 50(5), 1093-1102. doi: 10.1249/MSS.0000000000001532

Crouter, S. E., Clowers, K. G., & Bassett Jr, D. R. (2006). A novel method for using accelerometer data to predict energy expenditure. Journal of Applied Physiology, 100(4), 1324-1331.

Crouter, S. E., Kuffel, E., Haas, J. D., Frongillo, E. A., & Bassett Jr, D. R. (2010). Refined Two-Regression Model for the ActiGraph Accelerometer. Medicine and Science in Sports and Exercise, 42(5), 1029.

Crouter, S. E., Horton, M., & Bassett Jr, D. R. (2012). Use of a Two-regression model for estimating energy expenditure in children. Medicine and Science in Sports and Exercise, 44(6), 1177.

Examples



## Datasets

  data(count_data, package = "TwoRegression")
  data(all_data, package = "TwoRegression")

## Crouter 2006-2012 models

  TwoRegression(
    count_data, "Crouter 2006",
    movement_var = "Axis1", time_var = "time"
  )

  TwoRegression(
    count_data, "Crouter 2010",
    movement_var = "Axis1", time_var = "time"
  )

  TwoRegression(
    count_data, "Crouter 2012", movement_var = "Axis1",
    time_var = "time", model = "VA", check = FALSE
  )

  TwoRegression(
    count_data, "Crouter 2012", movement_var = "Vector.Magnitude",
    time_var = "time", model = "VM", check = FALSE
  )

## Hibbing 2018 models (can be vectorized)

  all_data$ENMO_CV10s <- NULL
  all_data$GVM_CV10s  <- NULL
  all_data$Direction  <- NULL

  result <- TwoRegression(
    all_data, "Hibbing 2018", gyro_var = "Gyroscope_VM_DegPerS",
    direction_var = "mean_magnetometer_direction",
    site = c("Left Ankle", "Right Ankle"), algorithm = 1:2
  )

  utils::head(result)

Two-regression-ready data frame

Description

A dataset with pre-processed primary accelerometer and IMU data that is ready for applying a two-regression algorithm.

Usage

all_data

Format

A data frame with 299 rows and 17 variables:

PID

Participant ID

file_source_PrimaryAccel

The filename of the primary accelerometer file

date_processed_PrimaryAccel

The date the primary accelerometer file was processed

file_source_IMU

The filename of the IMU file

date_processed_IMU

The date the IMU file was processed

Timestamp

The corresponding time for each row of data

day_of_year

The numeric day of the year, i.e., the Julian date

minute_of_day

The numeric minute of the day

ENMO

Euclidian Norm Minus One, in milli-g

Gyroscope_VM_DegPerS

Gyroscope vector magnitude, in degrees per second

mean_abs_Gyroscope_x_DegPerS

Rotation in x axis, degrees per second

mean_abs_Gyroscope_y_DegPerS

Rotation in y axis, degrees per second

mean_abs_Gyroscope_z_DegPerS

Rotation in z axis, degrees per second

mean_magnetometer_direction

Cardinal direction of magnetometer signal, averaged over one second

ENMO_CV10s

Coefficient of variation per 10-s, applied to Euclidian Norm Minus One

GVM_CV10s

Coefficient of variation per 10-s, applied to gyroscope vector magnitude

Direction

Direction changes per 5-s

Activity count data for demonstrating prior two-regression models

Description

A small amount of 10-s epoch activity counts for code examples

Usage

count_data

Format

A data frame with 30 rows and 5 variables:

time

POSIX. The timestamp

Axis1

numeric. The vertical axis activity counts

Axis2

numeric. The horizontal axis

Axis3

numeric. The lateral axis

Vector.Magnitude

numeric. The vector magnitude of all three axes

Calculate direction changes per five seconds in sliding windows

Description

get_directions is legacy code, whereas get_dcp5 is cleaner and (possibly) faster

Usage

get_directions(x, window_size = 5)

get_dcp5(x, window_size = 5)

Arguments

x

a numeric vector on which to perform the calculation

window_size

width of the sliding window, in data points

Value

a numeric vector of values, giving the number of direction changes in the sliding window that corresponds to each epoch of data

Develop a two-regression algorithm

Description

Develop a two-regression algorithm

Check if an object has class TwoRegression

Usage

fit_2rm(
  data,
  activity_var,
  sed_cp_activities,
  sed_activities,
  sed_cp_var,
  sed_METs,
  walkrun_activities,
  walkrun_cp_var,
  met_var,
  walkrun_formula,
  intermittent_formula,
  method = "user_unspecified"
)

is.TwoRegression(x)

Arguments

data

The data with which to develop the algorithm

activity_var

Character scalar. Name of the variable defining which activity is being performed

sed_cp_activities

Character vector. Activities to be included in the process of forming the sedentary classifier

sed_activities

Character vector. Actual sedentary activities

sed_cp_var

Character scalar. Name of the variable on which the sedentary cut-point is defined

sed_METs

Numeric scalar. Metabolic equivalent value to apply to sedentary activities

walkrun_activities

Character vector. Actual ambulatory activities

walkrun_cp_var

Character scalar. Name of the variable on which the walk/run cut-point is defined

met_var

Character scalar. Name of the variable giving actual energy expenditure (in metabolic equivalents)

walkrun_formula

Character scalar. Formula to use for developing the walk/run regression model

intermittent_formula

Character scalar. Formula to use for developing the intermittent activity regression model

method

character scalar. Optional name for the model, potentially useful for printing.

x

object to be tested

Value

An object of class 'TwoRegression'

See Also

predict.TwoRegression summary.TwoRegression plot.TwoRegression

Examples


set.seed(307)

data(all_data, package = "TwoRegression")
fake_sed <- c("Lying", "Sitting")
fake_lpa <- c("Sweeping", "Dusting")
fake_cwr <- c("Walking", "Running")
fake_ila <- c("Tennis", "Basketball")

fake_activities <- c(fake_sed, fake_lpa, fake_cwr, fake_ila)

all_data$Activity <- sample(fake_activities, nrow(all_data), TRUE)

all_data$fake_METs <- ifelse(
  all_data$Activity %in% c(fake_sed, fake_lpa),
  runif(nrow(all_data), 1, 2),
  runif(nrow(all_data), 2.5, 8)
)

fit_2rm(
  data = all_data,
  activity_var = "Activity",
  sed_cp_activities = c(fake_sed, fake_lpa),
  sed_activities = fake_sed,
  sed_cp_var = "ENMO",
  sed_METs = 1.25,
  walkrun_activities = fake_cwr,
  walkrun_cp_var = "ENMO_CV10s",
  met_var = "fake_METs",
  walkrun_formula = "fake_METs ~ ENMO",
  intermittent_formula = "fake_METs ~ ENMO + I(ENMO^2) + I(ENMO^3)"
)

Convert an object of class TwoRegression to a textual representation of the algorithm

Description

Convert an object of class TwoRegression to a textual representation of the algorithm

Usage

get_2rm_formula(object)

Arguments

object

the TwoRegression object

Develop a cut-point as part of the process for developing a two-regression algorithm

Description

Develop a cut-point as part of the process for developing a two-regression algorithm

Usage

get_cut_point(
  data,
  activity_var,
  included_activities,
  positive_activities,
  cp_var
)

Arguments

data

the data on which to perform the operation

activity_var

character scalar. Name of the variable that indicates what activity each data point corresponds to.

included_activities

character vector. All activities included in the data subset to be used for developing the cut-point

positive_activities

character vector. Activities considered part of the positive class

cp_var

Character scalar. Name of the variable on which the cut-point will be based.

Value

an object of class 'roc'

Coefficient of variation for two-regression models

Description

Calculates coefficient of variation using static or sliding methods, with potential for custom methods as well

Usage

cv_2rm(
  x,
  window_size = 10,
  approach = c("sliding", "static", "custom"),
  verbose = FALSE,
  ...
)

static_cv(x, window_size)

sliding_cv(x, window_size)

custom_cv(x, window_size, pad_size = window_size - 1, ...)

Arguments

x

a numeric vector on which to perform the calculation

window_size

width of the sliding window, in data points

approach

character scalar naming the desired calculation approach to use. Can be "sliding" (the default; see Crouter et al., 2010), "static" (see Crouter et al., 2006), or "custom" (see details below)

verbose

logical. Print progress updates?

...

arguments passed to functions in the RcppRoll package

pad_size

The number of NA values to append at the start and end of x (when approach == "custom") before executing the rolling minimum step.

Details

For approach == "sliding", the value for each epoch represents the lowest CV value of all the sliding windows that overlap with that epoch. For aproach == "static", a truncated vector of CV values is given, which reflects CV values from a non-overlapping sliding window. For approach == "custom", users can pass arguments into RcppRoll functions and create a variation on the sliding approach. Behavior of this feature is not well documented and subject to change if people start using it and requesting specific behavior.

Value

a numeric vector of values, giving the desired coefficient of variation

See Also

Crouter et al. (2010, Med Sci Sports Exerc) Crouter et al. (2006, J Appl Physiol)

Examples

data(raw_for_cv)
cv_2rm(raw_for_cv$ENMO)

Convert a linear model to a text representation of the prediction equation

Description

Convert a linear model to a text representation of the prediction equation

Usage

get_lm_formula(model)

Arguments

model

The linear model object

Apply a Hibbing 2018 two-regression algorithm

Description

Applies the specified two-regression algorithm from Hibbing et al. (2018, Med Sci Sports Exerc) to data from the primary accelerometer and IMU (if applicable)

Usage

apply_two_regression_hibbing18(
  which_algorithm = data.frame(Wear_Location = "Hip", Algorithm = 1),
  AG,
  verbose = FALSE,
  ...
)

Arguments

which_algorithm

a dataframe specifying which algorithm to use, based on Wear_Location and Algorithm columns

AG

data frame of ActiGraph data

verbose

logical. Print updates to console?

...

arguments passed to predict.TwoRegression

Value

a numeric vector of predicted energy expenditure values, expressed in metabolic equivalents

Internal functions for the 2RM wrapper

Description

Internal functions for the 2RM wrapper

Usage

TwoRegression_Hibbing18_variable_validate(
  varname,
  AG,
  algorithm,
  sensor = c("accelerometer", "gyroscope", "magnetometer")
)

TwoRegression_Hibbing18_algorithm_validate(algorithm)

Arguments

varname

character scalar. Name of variable to check

AG

data frame of ActiGraph data

algorithm

An integer/numeric scalar or vector giving the algorithm(s) to apply to the data from the primary accelerometer and (if applicable) IMU. Must be 1 (accelerometer only), 2 (accelerometer and gyroscope), 3 (accelerometer, gyroscope, and magnetometer), or any combination thereof

sensor

character scalar indicating if varname corresponds to an accelerometer variable or gyroscope variable

IMU data to check

Description

A dataset for demonstrating checks that are applied to IMU data.

Usage

imu_to_check

Format

A data frame with 300 rows and 8 variables:

file_source_IMU

The filename of the IMU file

date_processed_IMU

The date the IMU file was processed

Timestamp

The corresponding time for each row of data

Gyroscope_VM_DegPerS

Gyroscope vector magnitude, in degrees per second

mean_abs_Gyroscope_x_DegPerS

Rotation in x axis, degrees per second

mean_abs_Gyroscope_y_DegPerS

Rotation in y axis, degrees per second

mean_abs_Gyroscope_z_DegPerS

Rotation in z axis, degrees per second

mean_magnetometer_direction

Cardinal direction of magnetometer signal, averaged over one second

IMU data to collapse

Description

A partially-processed IMU dataset ready to be collapsed from raw samples to one-second summaries.

Usage

imu_to_collapse

Format

A data frame with 1500 rows and 17 variables:

Timestamp

The corresponding time for each row of data

Accelerometer.X

Secondary accelerometer x-axis data, in G

Accelerometer.Y

Secondary accelerometer y-axis data, in G

Accelerometer.Z

Secondary accelerometer z-axis data, in G

Temperature

Temperature of the IMU, in Celcius

Gyroscope.X

Gyroscope x-axis data, in degrees per second

Gyroscope.Y

Gyroscope y-axis data, in degrees per second

Gyroscope.Z

Gyroscope z-axis data, in degrees per second

Magnetometer.X

Magnetometer x-axis data, in micro-Teslas

Magnetometer.Y

Magnetometer y-axis data, in micro-Teslas

Magnetometer.Z

Magnetometer z-axis data, in micro-Teslas

file_source_IMU

The filename of the IMU file

date_processed_IMU

The date the IMU file was processed

ms

The millisecond value of the timestamp

mean_Accel_VM

Vector magnitude of the secondary accelerometer signal, in G

Gyroscope_VM_DegPerS

Gyroscope vector magnitude, in degrees per second

Magnetometer_VM_MicroT

Vector magnitude of the magnetometer signal, in micro-Teslas

Create summary plots for TwoRegression objects

Description

Four plots are generated: a threshold plot for both cut-points, and a model plot for both regression models

Usage

## S3 method for class 'TwoRegression'
plot(
  x = NULL,
  object = NULL,
  sed_cp_activities,
  sed_activities,
  sed_cpVar = NULL,
  activity_var,
  met_var,
  walkrun_activities,
  walkrun_cpVar,
  x_sed = NULL,
  y_sed = NULL,
  x_walkrun = NULL,
  y_walkrun = NULL,
  print = TRUE,
  ...
)

Arguments

x

passed from generic function but not used in the method

object

the TwoRegression object

sed_cp_activities

Character vector. Activities to be included in the process of forming the sedentary classifier

sed_activities

Character vector. Actual sedentary activities

sed_cpVar

character scalar. The name of the variable on which the cut-point is based

activity_var

Character scalar. Name of the variable defining which activity is being performed

met_var

character scalar. The name of the variable containing energy expenditure values, in metabolic equivalents

walkrun_activities

Character vector. Actual ambulatory activities

walkrun_cpVar

character scalar giving the name of the variable on which the walk/run cut-point is based

x_sed

numeric scalar giving x coordinate for label placement in sedentary cut-point plot

y_sed

numeric scalar giving y coordinate for label placement in sedentary cut-point plot

x_walkrun

numeric scalar giving x coordinate for label placement in walk/run cut-point plot

y_walkrun

numeric scalar giving y coordinate for label placement in walk/run cut-point plot

print

logical. Should the plot be arranged in a grid? If false, the panels will be returned in a list of gg/ggplot objects.

...

further arguments passed to plotting calls

Value

A two-by-two grid of summary plots

Examples

data(all_data, package = "TwoRegression")
all_data$PID <-
  rep(
    c("Test1", "Test2"),
    each = ceiling(nrow(all_data) / 2))[seq(nrow(all_data))]

fake_sed <- c("Lying", "Sitting")
fake_lpa <- c("Sweeping", "Dusting")
fake_cwr <- c("Walking", "Running")
fake_ila <- c("Tennis", "Basketball")

fake_activities <- c(fake_sed, fake_lpa, fake_cwr, fake_ila)

all_data$Activity <-
  sample(fake_activities, nrow(all_data), TRUE)

all_data$fake_METs <-
  ifelse(all_data$Activity %in% c(fake_sed, fake_lpa),
    runif(nrow(all_data), 1, 2),
    runif(nrow(all_data), 2.5, 8)
  )

ex_2rm <- fit_2rm(
  data = all_data,
  activity_var = "Activity",
  sed_cp_activities = c(fake_sed, fake_lpa),
  sed_activities = fake_sed,
  sed_cp_var = "ENMO",
  sed_METs = 1.25,
  walkrun_activities = fake_cwr,
  walkrun_cp_var = "ENMO_CV10s",
  met_var = "fake_METs",
  walkrun_formula = "fake_METs ~ ENMO",
  intermittent_formula = "fake_METs ~ ENMO + I(ENMO^2) + I(ENMO^3)"
)

model_plot_list <- plot(
  object = ex_2rm,
  sed_cp_activities = c(fake_sed, fake_lpa),
  sed_activities = fake_sed,
  sed_cpVar = "ENMO",
  activity_var = "Activity",
  met_var = "fake_METs",
  walkrun_activities = fake_cwr,
  walkrun_cpVar = "ENMO_CV10s",
  print = FALSE
)


  print(model_plot_list$sed_cut_point)
  print(model_plot_list$walkrun_cut_point)
  print(model_plot_list$walkrun_regression)
  print(model_plot_list$intermittent_regression)

  plot(
    object = ex_2rm,
    sed_cp_activities = c(fake_sed, fake_lpa),
    sed_activities = fake_sed,
    sed_cpVar = "ENMO",
    activity_var = "Activity",
    met_var = "fake_METs",
    walkrun_activities = fake_cwr,
    walkrun_cpVar = "ENMO_CV10s",
    print = TRUE
  )

Add a cut-point to a plot

Description

Add a cut-point to a plot

Usage

plotCP(fig, cutPoint, x = NULL, y = NULL)

Arguments

fig

A partially-assembled figure

cutPoint

The threshold value to insert into the figure

x

numeric scalar giving x coordinate for label placement

y

numeric scalar giving y coordinate for label placement

Plot behaviors for visualizing a sedentary cut-point

Description

Plot behaviors for visualizing a sedentary cut-point

Usage

plot_sed(
  object,
  sed_cp_activities,
  sed_activities,
  activity_var,
  sed_cpVar,
  met_var,
  ...
)

Arguments

object

TwoRegression object to plot

sed_cp_activities

activities to include in a subset of AG

sed_activities

Character vector. Actual sedentary activities

activity_var

Character scalar. Name of the variable defining which activity is being performed

sed_cpVar

character scalar. The name of the variable on which the cut-point is based

met_var

character scalar. The name of the variable containing energy expenditure values, in metabolic equivalents

...

additional arguments passed to plotCP

Plot behaviors for visualizing a walk/run cut-point

Description

Plot behaviors for visualizing a walk/run cut-point

Usage

plot_walkrun(
  object,
  activity_var,
  walkrun_activities,
  walkrun_cpVar,
  met_var,
  ...
)

Arguments

object

the TwoRegression object to plot from

activity_var

Character scalar. Name of the variable defining which activity is being performed

walkrun_activities

character vector giving values of activity_var that are walk/run activities

walkrun_cpVar

character scalar giving the name of the variable on which the walk/run cut-point is based

met_var

character scalar. The name of the variable containing energy expenditure values, in metabolic equivalents

...

further arguments passed to plotting calls

Predict metabolic equivalents from a TwoRegression object

Description

Predict metabolic equivalents from a TwoRegression object

Usage

## S3 method for class 'TwoRegression'
predict(
  object,
  newdata,
  min_mets = object$sed_METs,
  max_mets = 20,
  warn_high_low = TRUE,
  verbose = FALSE,
  ...
)

Arguments

object

the TwoRegression object

newdata

the data on which to predict metabolic equivalents (METs)

min_mets

the minimum allowable value for MET predictions. Defaults to the value stored in object$sed_METs

max_mets

the maximum allowable value for MET predictions. There is no value embedded in object. The default is 20

warn_high_low

logical. Issue warnings about values less than min_mets or greater than max_mets?

verbose

logical. Print processing updates?

...

further arguments passed to or from other methods

Value

A two-column data frame giving the activity classification (sedentary, walk/run, or intermittent activity) and the corresponding metabolic equivalent prediction

Examples

data(all_data, package = "TwoRegression")
all_data$PID <-
  rep(
    c("Test1", "Test2"),
    each = ceiling(nrow(all_data) / 2))[seq(nrow(all_data))]

train_data <- all_data[all_data$PID != "Test2", ]
test_data <- all_data[all_data$PID == "Test2", ]

fake_sed <- c("Lying", "Sitting")
fake_lpa <- c("Sweeping", "Dusting")
fake_cwr <- c("Walking", "Running")
fake_ila <- c("Tennis", "Basketball")

fake_activities <- c(fake_sed, fake_lpa, fake_cwr, fake_ila)

train_data$Activity <-
  sample(fake_activities, nrow(train_data), TRUE)

train_data$fake_METs <-
  ifelse(train_data$Activity %in% c(fake_sed, fake_lpa),
    runif(nrow(train_data), 1, 2),
    runif(nrow(train_data), 2.5, 8)
  )

ex_2rm <- fit_2rm(
  data = train_data,
  activity_var = "Activity",
  sed_cp_activities = c(fake_sed, fake_lpa),
  sed_activities = fake_sed,
  sed_cp_var = "ENMO",
  sed_METs = 1.25,
  walkrun_activities = fake_cwr,
  walkrun_cp_var = "ENMO_CV10s",
  met_var = "fake_METs",
  walkrun_formula = "fake_METs ~ ENMO",
  intermittent_formula = "fake_METs ~ ENMO + I(ENMO^2) + I(ENMO^3)"
)

predict(ex_2rm, test_data)

Primary accelerometer data to calculate coefficient of variation per 10-s

Description

A partially-processed primary accelerometer dataset ready to calculate the coefficient of variation per 10-s

Usage

raw_for_cv

Format

A data frame with 299 rows and 2 variables:

Block

A vestigial variable synonymous with row number

ENMO

Euclidian Norm Minus One, in milli-g

Primary accelerometer data to collapse

Description

A partially-processed primary accelerometer dataset ready to be collapsed from raw samples to one-second summaries.

Usage

raw_to_collapse

Format

A data frame with 24000 rows and 3 variables:

Accelerometer X

Primary accelerometer x-axis data, in G

Accelerometer Y

Primary accelerometer y-axis data, in G

Accelerometer Z

Primary accelerometer z-axis data, in G

Smooth two-regression estimates over specified periods

Description

Smooth two-regression estimates over specified periods

Usage

smooth_2rm(AG, time_var = "Timestamp", unit = "60 sec", verbose = FALSE, ...)

Arguments

AG

data frame of ActiGraph data

time_var

character scalar. Name of the timestamp variable (required to verify that input epoch length is 10 seconds)

unit

the interval to use for smoothing (see floor_date). Default is "60 sec"

verbose

logical. Print updates to console?

...

currently unused

Value

Smoothed data, collapsed in the specified intervals

Examples


data(all_data, package = "TwoRegression")

  result <- TwoRegression(
    all_data, "Hibbing 2018", gyro_var = "Gyroscope_VM_DegPerS",
    direction_var = "mean_magnetometer_direction",
    site = c("Left Ankle", "Right Ankle"), algorithm = 1:2
  )

  smooth_2rm(result)

Summary method for TwoRegression objects

Description

Summary method for TwoRegression objects

Usage

## S3 method for class 'TwoRegression'
summary(object, ...)

new_summary(object)

new_cut(
  z,
  object,
  element = c("sed_roc", "walkrun_roc"),
  label = c("sedentary", "walk_run")
)

new_model(
  z,
  object,
  element = c("walkrun_model", "intermittent_model"),
  label = c("walk_run", "intermittent_activity")
)

Arguments

object

The TwoRegression object to summarize

...

Arguments passed to DualCP_LOSO

z

An intermediate summary object for internal use

element

character. Element name to search for in z

label

character. A label for the cut point

Value

and object of class summary.TwoRegression, containing information about the cut-points and diagnostics, regression model performance, and overall algorithm performance during leave-one-participant-out-cross-validation.

Examples


set.seed(307)

data(all_data, package = "TwoRegression")
all_data$PID <- rep(
  c("Test1", "Test2"),
  each = ceiling(nrow(all_data) / 2)
)[seq(nrow(all_data))]

fake_sed <- c("Lying", "Sitting")
fake_lpa <- c("Sweeping", "Dusting")
fake_cwr <- c("Walking", "Running")
fake_ila <- c("Tennis", "Basketball")

fake_activities <- c(fake_sed, fake_lpa, fake_cwr, fake_ila)

all_data$Activity <-
  sample(fake_activities, nrow(all_data), TRUE)

all_data$fake_METs <-
  ifelse(all_data$Activity %in% c(fake_sed, fake_lpa),
    runif(nrow(all_data), 1, 2),
    runif(nrow(all_data), 2.5, 8)
  )

ex_2rm <- fit_2rm(
  data = all_data,
  activity_var = "Activity",
  sed_cp_activities = c(fake_sed, fake_lpa),
  sed_activities = fake_sed,
  sed_cp_var = "ENMO",
  sed_METs = 1.25,
  walkrun_activities = fake_cwr,
  walkrun_cp_var = "ENMO_CV10s",
  met_var = "fake_METs",
  walkrun_formula = "fake_METs ~ ENMO",
  intermittent_formula = "fake_METs ~ ENMO + I(ENMO^2) + I(ENMO^3)"
)

summary(
  ex_2rm,
  subject_var = "PID",
  MET_var = "fake_METs",
  activity_var = "Activity"
)