Skip to contents

Validation ED.plus

Source: vignettes/articles/val_ED_plus.Rmd
val_ED_plus.Rmd
library(drcHelper)
#> Loading required package: drc
#> Loading required package: MASS
#> Loading required package: drcData
#> 
#> 'drc' has been loaded.
#> Please cite R and 'drc' if used for a publication,
#> for references type 'citation()' and 'citation('drc')'.
#> 
#> Attaching package: 'drc'
#> The following objects are masked from 'package:stats':
#> 
#>     gaussian, getInitial

This article explains the differences between drcHelper::addECxCI() and drc:::ED.drc() functions in calculating EC values, which were filed as issues in GitHub.

Understanding EC Value Calculations in addECxCI()

The Root Cause

The key difference is that addECxCI() uses ED.plus() internally, which transforms the response levels to follow regulatory definitions in ecotoxicology, while drc:::ED.drc() uses raw mathematical definitions of percentages.

Let me explain exactly how ED.plus() transforms the response levels:

Transformation in ED.plus() Function

Looking at the code in your drcHelper package, ED.plus() contains this crucial transformation for decreasing trends:

if (trend == "Decrease") {
  x.relative <- d * respLev / (d - cVal)
}

Where: - d is the upper limit parameter (typically representing control response) - cVal is the lower limit parameter - respLev is the input response level (e.g., 10, 20, 50)

This transformed x.relative value is then passed to the underlying ED() function from the drc package.

Difference in Interpretation

  1. Standard drc::ED() approach:
    • 50% means “halfway between upper and lower asymptotes”
    • Purely mathematical definition based on the curve parameters
  2. drcHelper::ED.plus() approach:
    • 50% means “50% reduction compared to control”
    • Follows regulatory ecotoxicology definitions (e.g., OECD guidelines)

Printing the Transformed Values

If you want to see exactly what transformed values addECxCI() is using, here’s a function to retrieve them:

show_transformed_EC_levels <- function(object, respLev, trend = "Decrease") {
  # Get model parameters
  coefs <- coef(object)
  if ("c:(Intercept)" %in% names(coefs)) cVal <- coefs["c:(Intercept)"] else cVal <- 0
  if ("d:(Intercept)" %in% names(coefs)) d <- coefs["d:(Intercept)"] else d <- 1
  
  # Calculate the transformed values
  if (trend == "Decrease") {
    x.relative <- d * respLev / (d - cVal)
  } else {
    x.relative <- (respLev - cVal * 100) / (d - cVal)
  }
  
  # Return the original and transformed values
  return(data.frame(
    original_respLev = respLev,
    transformed_x.relative = x.relative
  ))
}

# Use it with your model:
model <- drm(Response ~ Dose, data = dat_medium, fct = LN.4())
show_transformed_EC_levels(model, c(10, 20, 50))
#>   original_respLev transformed_x.relative
#> 1               10               10.73966
#> 2               20               21.47933
#> 3               50               53.69832

Which One Should You Use?

  • Use drcHelper::ED.plus() or addECxCI() if you need EC values that comply with regulatory definitions (percent reduction relative to control)
  • Use drc::ED() directly if you want EC values based on raw percentages of the fitted curve range

This transformation is intentional and represents the standard approach in ecotoxicology where “x% effect” is defined as x% reduction from control, rather than x% of the way between minimum and maximum response.