R/simulationBiasMSE.r

In socale/frailtypack: Shared, Joint (Generalized) Frailty Models; Surrogate Endpoints

# Function used to summarize the simulation results. results in this case just include Biase and MSE
# param.estim is the dataset of the estimates from the simulation study
simulationBiasMSE <- function(param.estim, R2 = 0.81, ktau = 0.378, nb.simul, dec = 3, CP = 0, cpR, cpT){
  
  # recherche des lignes correspondantes aux simulations qui n'ont pas convergees (concerne le cas des programmes esimees par MPI-OpenMP)
  somme_row <- NULL
  for(i in 1:nrow(param.estim))
    somme_row[i] <- sum(param.estim[i,])
  
  #on les exclu du dataframe
  donnee <- param.estim[somme_row!=0,]
  
  
  d <- data.frame(parametre = c("R2_adj","Ktau","R:n(%)"))
  d$True <- c(R2, ktau, "")
  d$Mean <- NA
  d$Biais <- NA
  d$MSE <- NA
  if(CP == 1) d$CP <- NA
  #Ktau
  m_ktau <- mean(donnee$tau[!is.na(donnee$tau)])
  b_tau <- ktau-m_ktau # biais Ktau
  mse_tau <- mean((ktau-donnee$tau[!is.na(donnee$tau)])^2) # MSE
                
  #R2_adj
  m_r2_ad <- mean(donnee$R2trial[!is.na(donnee$R2trial)])
  b_r2_ad <- R2-m_r2_ad # biais Ktau
  mse_r2_ad <- mean((R2-donnee$R2trial[!is.na(donnee$R2trial)])^2) # MSE
                
  #NA
  n_NA <- nb.simul - nrow(donnee)
          
  #sauvagarde dans d
  if(CP == 0){
    d[1,3:5] <- c(round(m_r2_ad, dec),round(b_r2_ad, dec),round(mse_r2_ad, dec))
    d[2,3:5] <- c(round(m_ktau, dec),round(b_tau, dec),round(mse_tau, dec))
    d[3,3:5] <- c("",paste(n_NA,"(",round(100*n_NA/nb.simul),")",sep = ""),"")
  }else{
    d[1,3:6] <- c(round(m_r2_ad, dec),round(b_r2_ad, dec),round(mse_r2_ad, dec), cpR)
    d[2,3:6] <- c(round(m_ktau, dec),round(b_tau, dec),round(mse_tau, dec), cpT)
    d[3,3:6] <- c("",paste(n_NA,"(",round(100*n_NA/nb.simul),")",sep = ""),"","")
  }

  return(d)
}