Robust Calculation of Model-Averaged Effective Doses

This function serves as a robust wrapper for drc::maED. It calculates model-averaged effective doses (EDs) for specified response levels. The key feature is its resilience to errors; it iterates through each response level individually and handles failures gracefully by returning NA values for that level, rather than terminating the entire operation.

Usage

maED_robust(
  mod,
  fct_ls = NULL,
  respLev = c(10, 20, 50),
  interval = "buckland",
  CI_level = 0.95,
  verbose = FALSE,
  ...
)

Arguments

mod: A model object of class 'drc', which serves as the base model for the averaging.
fct_ls: A list of alternative dose-response functions (e.g., LL.3(), W1.4()) to be used in the model averaging process. The list should be named.
respLev: A numeric vector specifying the response levels (in percentages) for which to calculate the EDs (e.g., c(10, 50) for EC10 and EC50).
interval: A character string specifying the type of confidence interval to be supplied. The default is "buckland". See drc::maED for other options.
CI_level: A numeric value between 0 and 1 specifying the confidence level for the confidence intervals. Default is 0.95.
verbose: A logical value. If TRUE, the function will print status messages about the calculation progress and any errors encountered for each response level. Default is FALSE.
...: Additional arguments to be passed to the underlying drc::maED function.

Value

A data.frame with one row for each response level specified in respLev. The columns are:

Estimate: The estimated model-averaged effective dose.
stderr: The standard error of the estimate.
Lower: The lower bound of the confidence interval.
Upper: The upper bound of the confidence interval.
confint_level: The confidence level used for the interval.
confint_method: The method used for the confidence interval calculation.
model: A character string listing the models used for averaging.
EC: The response level (as a percentage).

If the calculation for a specific response level fails or results in a non-positive estimate, the corresponding row will contain NA values for Estimate, stderr, Lower, and Upper.

Details

The function enhances drc::maED by introducing a robust calculation loop. It iterates over each element of respLev and calls drc::maED within a tryCatch block. This approach isolates failures, preventing an error at one response level (e.g., an EC99 that cannot be estimated) from halting the calculation of others.

Furthermore, after a successful calculation, the function checks if the resulting 'Estimate' is positive. If the estimate is NA, non-positive, or if the tryCatch block catches an error, the function returns a structured row of NAs for that response level, ensuring a consistent output format.

Author

Hannes Reinwald

Examples

data(lettuce)
base_model <- drm(weight ~ conc, data = lettuce, fct = BC.5())
model_list <- list(W2.4 = W2.4())
maED_robust(base_model, fct_ls = model_list, respLev = c(10, 50))
#>     Estimate    stderr      Lower    Upper confint_level confint_method
#>        <num>     <num>      <num>    <num>         <num>         <char>
#> 1:  3.561851  1.610667   0.405001  6.71870          0.95       buckland
#> 2: 11.952400 11.849870 -11.272918 35.17772          0.95       buckland
#>        model    EC
#>       <char> <num>
#> 1: BC.5/W2.4    10
#> 2: BC.5/W2.4    50