In mladenjovanovic/dorem: Dose Response Modeling

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  fig.path = "man/figures/README-",
  fig.align = "center",
  fig.width = 6,
  fig.height = 4,
  dpi = 300,
  out.width = "90%",
  auto_pdf = TRUE,
  message = FALSE,
  warning = FALSE
)

dorem

The goal of dorem is to provide easy-to-use dose-response models utilized in sport science. This package is currently in active development phases.

Installation

You can install the development version from GitHub with:

# install.packages("devtools")
devtools::install_github("mladenjovanovic/dorem")

require(dorem)

Example

To provide very simplistic example of dorem, I will use example data provided in supplementary material of Clarke & Skiba, 2013 paper, freely available on the publisher website. Data set contains cycling training load (i.e. dose) measured using the BikeScore metric (in AU) over 165 days, with occasional training response measured using 5-min Power Test (in Watts). Banister model (explained in aforementioned paper) is applied to understand relationship between training dose (i.e., BikeScore metric) and training response (i.e., 5-min Power Test):

require(dorem)
require(tidyverse)
require(cowplot)

data("bike_score")

banister_model <- dorem(
   Test_5min_Power ~ BikeScore,
   bike_score,
   method = "banister"
)

# Print results
banister_model

# get coefs
coef(banister_model)

# Get model predictions
bike_score$pred <- predict(banister_model, bike_score)$.pred

# Plot
dose <- ggplot(bike_score, aes(x = Day, y = BikeScore)) +
  theme_cowplot(10) +
  geom_bar(stat = "identity") +
  xlab(NULL)

response <- ggplot(bike_score, aes(x = Day, y = pred)) +
   theme_cowplot(10) +
   geom_line() +
   geom_point(aes(y = Test_5min_Power), color = "red") +
   ylab("Test 5min Power")

cowplot::plot_grid(dose, response, ncol = 1)

dorem also allows more control and setup using the control parameter. In the next example, cross-validation of 3 repeats and 5 folds will be performed, with additional feature of shuffling the predictors and evaluating how the model predicts on random predictors (i.e., dose):

banister_model <- dorem(
   Test_5min_Power ~ BikeScore,
   bike_score,
   method = "banister",

   # control setup
   control = dorem_control(
    shuffle = TRUE,
    optim_method = "L-BFGS-B",
    optim_maxit = 1000,
    cv_folds = 3,
    cv_repeats = 5
   )
)   

banister_model

To plot model predictions, including the CV as gray area and shuffle as dotted line, use:

plot(banister_model, type = "pred") + theme_minimal()

To plot model coefficients across CV folds:

plot(banister_model, type = "coef") + theme_minimal()

To plot model performance across CV folds (i.e., training and testing folds):

plot(banister_model, type = "perf") + theme_minimal()

References

Clarke DC, Skiba PF. 2013. Rationale and resources for teaching the mathematical modeling of athletic training and performance. Advances in Physiology Education 37:134–152. DOI: 10.1152/advan.00078.2011.

mladenjovanovic/dorem documentation built on July 23, 2022, 7:12 a.m.