Skip to contents

Finite-difference sensitivity for dynamic predictions

Source: R/dynamic_sensitivity.R
dynamic_sensitivity.Rd

dynamic_sensitivity() perturbs parameters in a dynamic prediction and returns unscaled and scaled sensitivity coefficients. It is designed as a lightweight diagnostic for sampling design and parameter identifiability.

Usage

dynamic_sensitivity(
  type = c("growth", "inactivation"),
  profile,
  start,
  parameters = NULL,
  relative_delta = 1e-06,
  times = NULL,
  ...
)

Arguments

type

Character string. One of "growth" or "inactivation".

profile

A dynamic_profile() object or compatible data frame.

start

Named list of parameter values passed to the dynamic prediction function.

parameters

Character vector of parameter names to perturb. Defaults to all numeric scalar entries in start.

relative_delta

Relative perturbation size.

times

Optional output times.

...

Additional arguments passed to predict_dynamic_growth() or predict_dynamic_inactivation().

Value

A data frame with time, parameter, prediction, sensitivity, and scaled sensitivity.

Examples

profile <- dynamic_profile(time = c(0, 5, 10), temperature = c(10, 15, 20))
sens <- dynamic_sensitivity(
  "growth",
  profile = profile,
  start = list(logN0 = 2, logNmax = 8, a = 0.08, Tmin = 7, lag = 1),
  times = seq(0, 10, by = 2),
  dt = 0.25
)
head(sens)
#>   time parameter prediction sensitivity scaled_sensitivity
#> 1    0     logN0   2.000000   1.0000000           2.000000
#> 2    2     logN0   2.025812   0.9999999           2.000000
#> 3    4     logN0   2.107644   0.9999997           1.999999
#> 4    6     logN0   2.233674   0.9999993           1.999999
#> 5    8     logN0   2.409684   0.9999984           1.999997
#> 6   10     logN0   2.640966   0.9999966           1.999993