#################################################################################################
### A function to generate simulated data sets and compute the p-values using our new method ####
#################################################################################################

# Input:
# nrun: number of simulated data sets
# n: number of observations in the data set
# p: number of molecular variables
# q: number of clinical covariates
# muX: shift for the relevant molecular variables
# muz: shift for the clinical covariates
# prel: number of relevant molecular variables
# nperm: number of permutation used to compute the p-value
# mstop: a vector giving the values of mstop that have to be considered
# XeqZ: should the relevant X variables be equal to the Z variables?

# Output:
# a nrun x (length(mstop)+2) matrix with the p-values obtained with different values of mstop and with the globaltest (last column)

simulation<-function(nrun,n=100,p=1000,q=5,muX,muZ,prel,nperm,mstop,XeqZ=FALSE)
{

result<-matrix(0,nrun,length(mstop)+2)
goeman<-function(X,Y,Z)
{
globaltest(X=t(X),Y=Y,adjust=Z)
}


for (i in 1:nrun)
 {
 print(paste("i=",i))
 set.seed(i)
 Y<-as.factor(rbinom(n,1,0.5))
 X<-as.data.frame(matrix(rnorm(n*p),n,p))
 if (prel>0 & muX>0)
  {
  X[Y==1,1:prel]<-X[Y==1,1:prel]+muX
  }

 if (!XeqZ)
  {
  Z<-as.data.frame(matrix(rnorm(n*q),n,q))
  if (muZ>0)
   {
   Z[Y==1,]<-Z[Y==1,]+muZ
   }
  }
 else
  {
  Z<-X[,1:prel]
  }

 my.testi<-globalboosttest(X,Y,Z,nperm=nperm,mstop=mstop,mstopAIC=TRUE,plot=TRUE,pvalueonly=TRUE)
 result[i,1:(length(mstop)+1)]<-my.testi$pvalue
 result[i,length(mstop)+2]<-p.value(goeman(X,Y,Z))
 }

colnames(result)<-c(paste("mstop=",mstop),"AIC","globaltest")
rownames(result)<-paste("run",1:nrun,sep="")

return(result)

}