functions-define-truth-logistic.R

# Reference: Julie Dudášová, Zdeněk Valenta, Jeffrey R. Sachs (2024).
# "Elucidating vaccine efficacy using a correlate of protection, demographics, and logistic regression"
# This program is released under the GNU GPL3 license. Copyright © 2024 Merck & Co., Inc., Rahway, NJ, USA and its affiliates. All rights reserved.

defineTruth <- function() {
  variable <- list()
  parameter <- list()
  vaccinated <- list()
  control <- list()
  
  ## Name ----
  name <- "results-hill-effect30-n15000"
  
  ## PoD model for data generation ----
  PoD <- "I(ifelse(var.titer > 0, param.pmax*(param.pmax.covariateBinary^var.covariateBinary) /(
  1+(var.titer/(param.et50*(param.et50.covariateBinary^var.covariateBinary))
  )^(param.gamma*(param.gamma.covariateBinary^var.covariateBinary) )), param.pmax))"
  
  # Independent variables
  variable$name <- c("var.titer", "var.covariateBinary")
  
  # Parameters
  parameter$name <- c("param.pmax",
                      "param.pmax.covariateBinary",
                      "param.et50",
                      "param.et50.covariateBinary",
                      "param.gamma",
                      "param.gamma.covariateBinary")
  parameter$value <- list(param.pmax = 0.033,
                          param.pmax.covariateBinary = 1,
                          param.et50 = 7.12,
                          param.et50.covariateBinary = 1.361, #1.131; 1.361 # for 10%; 30% effect
                          param.gamma = 7,
                          param.gamma.covariateBinary = 1)
  
  ## Populations ----
  
  # Vaccinated 
  
  # titer distribution
  vaccinated$N = 10000
  vaccinated$mean = 10
  vaccinated$stdDev = 2
  
  # covariate information
  
  # the first value in the interval will be adjusted in the code to 1, the other into 0 
  vaccinated$covariateInfo$var.covariateBinary <- list("name" = "var.covariateBinary",
                                                  "values" = c("valueA","valueB"),
                                                  "prob" = c(0.25, 0.75))
  
  # Control
  # titer distribution
  control$N = 5000
  control$mean = 5
  control$stdDev = 2
  
  # covariate information
  control$covariateInfo$var.covariateBinary <- list("name" = "var.covariateBinary",
                                                       "values" = c("valueA","valueB"),
                                                       "prob" = c(0.25, 0.75))
  
  
  ## Immunogenicity subsetting methods: "Full", "Fixed" ----
  method <- list(name = "Full",
                 value = NA)

  # method <- list(name = "Fixed",
  #                value = 1600)
  
  PoDModel <- MakeModelObj(PoD,
                           variable$name,
                           variable$range,
                           parameter$name,
                           parameter$range,
                           parameter$value)
  
  return(list(
    name = name,
    vaccinated = vaccinated,
    control = control,
    PoDModel = PoDModel,
    method = method,
    adjustTiters = FALSE, # this titer adjustment improves convergence:
    adjustFrom = 0, # titers shouldn't be negative, otherwise the negLL fails sometimes
    adjustTo = 0
  ))
}