Perform local sensitivity analysis simulations — sg_localsens

Generates simulation datasets for a model varying each specified parameter within provided bounds or relative percentage. Supports multiple outputs and covariates.

Usage

sg_localsens_sim(
  model,
  params,
  stimes,
  output = NULL,
  perc = 0.2,
  lb = NULL,
  ub = NULL,
  cov = NULL,
  et,
  theta = NULL,
  n_sim = 10
)

Arguments

model: GMO. RxODE2 model object.
params: Character vector of parameter names to vary in the analysis.
stimes: Numeric vector of time points for the simulation.
output: Character vector of model outputs (variables) to keep. Default is NULL (all outputs).
perc: Numeric. If lb/ub not provided, defines ±percentage range around parameter base value. Default 0.2.
lb: Numeric vector of lower bounds for each parameter. Length must match params. Default NULL.
ub: Numeric vector of upper bounds for each parameter. Length must match params. Default NULL.
cov: Character vector of covariate names to include in the simulation. Default NULL.
et: Event table for the simulation.
theta: Named numeric vector of baseline parameters. Default NULL.
n_sim: Integer. Number of points to simulate between LB and UB for each parameter. Default 10.

Value

A tibble containing the simulated results with the following columns:

NPOP: Optional. Population identification number. From simulation.
ID: Optional. Individual subject ID. From simulation.
TIME: Time points of simulation.
VAR: Name of the output variable (model compartment or biomarker).
VALUE: Simulated value of the variable.
mean, median, min, max, sd, etc.: Optional. Aggregated statistics, if aggregation applied.
COV1,...COVn: Optional. Covariate values, if cov provided.
PARNAME: Name of the parameter being varied in this simulation.
PARVAL: Value of the parameter used in this simulation.
PARVAL_NORM: Normalized parameter value between 0 and 1 relative to LB and UB. Useful for plotting color gradients.

Examples

# \donttest{
library(rxode2)
library(tibble)
mod_ex <- RxODE({
  # Doses in mg
  # Time in hours
  ### Parameter values
  # Typical
  POPCL = 5;
  POPVC = 180;
  POPQ = 7;
  POPVP = 52;
  POPKTR = 6;
  FBIOPAR = 1;
  # Covariate coefficients
  POPIGFRCOV = 0.8;
  POPPATCOVCLGR1 = 0.85;
  POPPATCOVCLGR2 = 0.85;
  POPPATCOVCLGR3 = 0.85;
  POPPATCOVCLGR4 = 0.85;
  # Random effects
  PPVCL = 0;
  PPVVC = 0;
  PPVKTR = 0;
  # Residual error
  RUV = 0;
  ### Covariates
  PATCOVCL = 1
  if(POPN == 1){PATCOVCL = POPPATCOVCLGR1}
  if(POPN == 2){PATCOVCL = POPPATCOVCLGR2}
  if(POPN == 3){PATCOVCL = POPPATCOVCLGR3}
  if(POPN == 4){PATCOVCL = POPPATCOVCLGR4}
  IGFRCOV = (IGFR/112)^POPIGFRCOV
  ## Parameters
  CL = POPCL * IGFRCOV * PATCOVCL * exp(PPVCL);
  VC = POPVC * exp(PPVVC);
  Q = POPQ;
  VP = POPVP;
  KTR = POPKTR * exp(PPVKTR);
  ### Explicit functions
  Cc = Ac/VC;                 # nmol/L
  Cp = Ap/VP;                 # nmol/L
  ### Initial conditions
  At1(0) = 0;         # mg
  At2(0) = 0;         # mg
  At3(0) = 0;         # mg
  At4(0) = 0;         # mg
  At5(0) = 0;         # mg
  At6(0) = 0;         # mg
  Ad(0) = 0;          # mg
  Ac(0) = 0;          # mg
  Ap(0) = 0;          # mg
  ### ODEs
  d/dt(At1) = - KTR*At1;
  d/dt(At2) = KTR*At1 - KTR*At2;
  d/dt(At3) = KTR*At2 - KTR*At3;
  d/dt(At4) = KTR*At3 - KTR*At4;
  d/dt(At5) = KTR*At4 - KTR*At5;
  d/dt(At6) = KTR*At5 - KTR*At6;
  d/dt(Ad) = KTR*At6 - KTR*Ad;
  d/dt(Ac) = KTR*Ad - CL*Cc - Q*(Cc - Cp);
  d/dt(Ap) = Q*(Cc - Cp);
  FBIO = FBIOPAR
  f(At1) = FBIO*1000000/505;
  CHECKRUV = RUV;
  Cc_ResErr = Cc + RUV*Cc;
})
#>  
#>  
et_base <- tribble(
  ~id, ~time, ~evid, ~cmt, ~amt, ~addl, ~ii, ~IGFR, ~POPN,
  1,   0,     1,     1,    10,   2,     24,  112,   1
)
sens_loc <- sg_localsens_sim(
  model = mod_ex,
  params = c("POPCL", "POPVC"),
  stimes = seq(0, 168, 0.1),
  output = "Cc",
  perc = 0.9,
  cov = c("IGFR", "POPN"),
  et = et_base
)
# }