Skip to contents

Calculate ACWR using EWMA Method with Confidence Bands

Source: R/calculate_acwr_ewma.R
calculate_acwr_ewma.Rd

Calculates the Acute:Chronic Workload Ratio (ACWR) using Exponentially Weighted Moving Average (EWMA) with optional bootstrap confidence bands.

Usage

calculate_acwr_ewma(
  activities_data,
  activity_type = NULL,
  load_metric = "duration_mins",
  method = c("ra", "ewma"),
  acute_period = 7,
  chronic_period = 28,
  half_life_acute = 3.5,
  half_life_chronic = 14,
  start_date = NULL,
  end_date = NULL,
  user_ftp = NULL,
  user_max_hr = NULL,
  user_resting_hr = NULL,
  smoothing_period = 7,
  ci = FALSE,
  B = 200,
  block_len = 7,
  conf_level = 0.95
)

Arguments

activities_data

A data frame of activities from load_local_activities().

activity_type

Optional. Filter activities by type. Default NULL includes all.

load_metric

Method for calculating daily load. Default "duration_mins".

method

ACWR calculation method: "ra" (rolling average) or "ewma". Default "ra".

acute_period

Days for acute window (for RA method). Default 7.

chronic_period

Days for chronic window (for RA method). Default 28.

half_life_acute

Half-life for acute EWMA in days. Default 3.5.

half_life_chronic

Half-life for chronic EWMA in days. Default 14.

start_date

Optional. Analysis start date. Defaults to one year ago.

end_date

Optional. Analysis end date. Defaults to today.

user_ftp

Required if load_metric = "tss".

user_max_hr

Required if load_metric = "hrss".

user_resting_hr

Required if load_metric = "hrss".

smoothing_period

Days for smoothing ACWR. Default 7.

ci

Logical. Whether to calculate confidence bands (EWMA only). Default FALSE.

B

Number of bootstrap iterations (if ci = TRUE). Default 200.

block_len

Block length for moving-block bootstrap (days). Default 7.

conf_level

Confidence level (0-1). Default 0.95 (95\% CI).

Value

A data frame with columns: date, atl, ctl, acwr, acwr_smooth, and if ci = TRUE and method = "ewma": acwr_lower, acwr_upper.

Details

This function extends the basic ACWR calculation with two methods:

  • RA (Rolling Average): Traditional rolling mean approach (default).

  • EWMA (Exponentially Weighted Moving Average): Uses exponential decay with configurable half-lives. More responsive to recent changes.

EWMA Formula: The smoothing parameter alpha is calculated from half-life: alpha = ln(2) / half_life. The EWMA update is: E_t = alpha * L_t + (1-alpha) * E_{t-1} where L_t is daily load and E_t is the exponentially weighted average.

Confidence Bands: When ci = TRUE and method = "ewma", uses moving-block bootstrap to estimate uncertainty. The daily load sequence is resampled in weekly blocks (preserving within-week correlation), ACWR is recalculated, and percentiles form the confidence bands. This accounts for temporal correlation in training load patterns.

Examples

# Example using pre-calculated sample data
data("sample_acwr", package = "Athlytics")
head(sample_acwr)
#> # A tibble: 6 × 5
#>   date         atl   ctl  acwr acwr_smooth
#>   <date>     <dbl> <dbl> <dbl>       <dbl>
#> 1 2023-02-03  32.6  39.0 0.837       0.888
#> 2 2023-02-04  30.6  38.6 0.793       0.881
#> 3 2023-02-05  31.3  38.2 0.819       0.883
#> 4 2023-02-06  31.6  37.8 0.835       0.867
#> 5 2023-02-07  35.9  38.6 0.93        0.863
#> 6 2023-02-08  37.9  38.1 0.994       0.866

if (FALSE) { # \dontrun{
# Full workflow with real data - Load local activities
activities <- load_local_activities("export_12345678.zip")

# Calculate ACWR using Rolling Average (RA)
acwr_ra <- calculate_acwr_ewma(activities, method = "ra")

# Calculate ACWR using EWMA with confidence bands
acwr_ewma <- calculate_acwr_ewma(
  activities,
  method = "ewma",
  half_life_acute = 3.5,
  half_life_chronic = 14,
  ci = TRUE,
  B = 200
)

# Compare both methods
head(acwr_ewma)
} # }