asVPC: Average Shifted Visual Predictive Checks

#' draw the original visual predictive check plot
#' 
#' @import ggplot2 plyr
#' @title Visual predictive checks
#' @param orig.data NONMEM data 
#' @param sim.data simulated data from NONMEM
#' @param N.timebin number of time bin
#' @param N.sim number of simulation
#' @param q.list list of quantiles for VPC plot
#' @param alpha significance level of CI for each quantile
#' @param X.name x label in VPC plot
#' @param Y.name y label in VPC plot
#' @param main.title title of plot
#' @param opt.DV.point option for drawing data points
#' @param opt.DV.quantile.line option for drawing quantiles of the original data
#' @param opt.SIM.quantile.line option for drawing quantiles of simulated data
#' @param opt.SIM.quantile.CI.area option for drawing confidence area of 
#'                                  quantiles for simulated data
#' @param Y.min minimum of y axis in VPC plot
#' @param Y.max maximum of y axis in VPC plot
#' @param plot.flag draw plot if TRUE; generate data for drawing plot if FALSE
#' @return plot or the values to draw plot
#' @export
#' @examples
#' data(origdata)
#' data(simdata)
#' VPC.graph.CI(origdata,simdata,10,100)

VPC.graph.CI<-function(orig.data,sim.data,N.timebin=NULL,N.sim,
                    q.list=c(0.05,0.5,0.95),
                    alpha=0.05,
                    X.name="TIME",Y.name="DV", ID.name="ID",
                    main.title=NULL,
                    opt.DV.point=FALSE,
                    opt.DV.quantile.line=TRUE,
                    opt.SIM.quantile.line=FALSE,
                    opt.SIM.quantile.CI.area=TRUE,
                    Y.min=NULL,Y.max=NULL,
                    X.min=NULL,X.max=NULL,                   
                    plot.flag=TRUE,
                    Y0.adjust=TRUE,
                    maxK=NULL,beta=0.2,lambda=0.3,
                    CIarea.box=FALSE,
                    bin.grid=TRUE,orig.flag=TRUE,...){
   SIM.CIarea.1<-NULL
   SIM.CIarea.2<-NULL
   SIM.CIarea.3<-NULL
   DV.point<-NULL
   DV.quant<-NULL
   SIM.quant<-NULL 
   ID<-NULL;G<-NULL

   optK<-function(orig.data,X.name="TIME",Y.name="DV",ID.name="ID",
                  maxK=NULL,beta=0.2,lambda=0.3){
     plot.data<-data.frame(X=orig.data[,X.name],
                           Y=orig.data[,Y.name],ID=orig.data[,ID.name])
     if(is.null(maxK)) 
       maxK <- max(c(round(nrow(orig.data)/10),
                     tapply(plot.data$X,plot.data$ID,length)))
     keep.tot<-NULL
     for(k in 2:maxK){
       time.bin<-makeCOVbin(plot.data$X,k)  
       J.temp<-sum((tapply(plot.data$X,time.bin$COV.bin,var)^beta)*
                     table(time.bin$COV.bin))
       temp.U<-ifelse(J.temp!=0,log(J.temp),0)+lambda*beta*k
       keep.tot<-rbind(keep.tot,c(k,temp.U))
     }
     colnames(keep.tot)<-c("k","U")
     keep.tot<-data.frame(keep.tot)
     return(keep.tot$k[which(keep.tot$U==min(keep.tot$U))])
   }
   
   findQuantile<-function(DV.data,TIME.data,time.bin.temp,q){
      TIME<-NULL
      DV.quantile<-NULL
      for(i in 1:nrow(time.bin.temp$COV.bin.summary)){
         if(time.bin.temp$COV.bin.summary$n.bin[i]!=0){
            sel.id<-which(time.bin.temp$COV.bin==
                               time.bin.temp$COV.bin.summary[i,1])
            DV.temp<-c(as.matrix(DV.data)[sel.id,])
            DV.quantile<-rbind(DV.quantile,quantile(DV.temp,prob=q,na.rm=TRUE))        
         } else{
            DV.quantile<-rbind(DV.quantile,NA) 
         }
      }
      DV.q<-data.frame(time.bin.temp$COV.bin.summary,DV.quantile)
      colnames(DV.q)[-(1:(ncol(DV.q)-length(q)))]<-
                                            paste("Q",round(q*100),"th",sep="")
      return(DV.q)
   }
  
   findQuantileCI<-function(sim.data,TIME.data,time.bin.temp,q,alpha,N.sim){
      orig.sim.Q.temp<-apply(sim.data,2,function(x) 
      findQuantile(x,TIME.data,time.bin.temp,q)[,
                              -(1:ncol(time.bin.temp$COV.bin.summary))])
      keep.name<-NULL#names(orig.sim.Q.temp[[1]])
      orig.sim.Q.temp<-array(unlist(orig.sim.Q.temp),
                             dim=c(nrow(time.bin.temp$COV.bin.summary),
                                   length(q),N.sim))
      Q.CI<-list()
      for(i in 1:length(q)){
         Q.CI[[i]]<-cbind(findQuantile(sim.data,TIME.data,
                                       time.bin.temp,q[i])[,4],
                          t(apply(orig.sim.Q.temp[,i,],1,
                           function(x) quantile(x,
                                               probs=c(alpha/2,1/2,1-alpha/2),
                                               na.rm=TRUE))))
         rownames(Q.CI[[i]])<-time.bin.temp$COV.bin.summary[,1]    
         keep.name<-c(keep.name,paste("Q",round(q[i]*100),"th",sep=""))
         colnames(Q.CI[[i]])[1]<-keep.name[i]
      }  
      names(Q.CI)<-keep.name
      return(list(TIME.bin.summary=time.bin.temp$COV.bin.summary,Q.CI=Q.CI))
   } 
   
   if(is.null(N.timebin)) 
     N.timebin<-optK(orig.data,X.name,Y.name,ID.name,maxK,beta,lambda)
   if(is.null(main.title))
     main.title<-paste("N.timebin=",N.timebin,sep="")
   
   
   plot.data<-data.frame(orig.data,X=orig.data[,X.name],Y=orig.data[,Y.name])
   time.bin<-makeCOVbin(plot.data$X,N.timebin)   
   if(is.null(Y.min)) Y.min<-min(c(unlist(sim.data),plot.data$Y),na.rm=T)
   if(is.null(Y.max)) Y.max<-max(c(unlist(sim.data),plot.data$Y),na.rm=T)

   if(is.null(X.min)) X.min<-min(c(plot.data$X,Q.CI[[1]][1,2]),
                                 na.rm=T)-range(plot.data$X)[2]/20
   if(is.null(X.max)) X.max<-max(c(plot.data$X,Q.CI[[1]][bintot.N,3]),
                                 na.rm=T)+range(plot.data$X)[2]/20
   
   P.temp<-ggplot(plot.data,aes(x=X,y=Y))+ylim(Y.min,Y.max)+xlim(X.min,X.max)+
     
                 labs(x=X.name,y=Y.name,title=main.title)
   if(bin.grid){
     temp.tick<-unique(unlist(time.bin$COV.bin.summary[,4:5]))
     P.temp<-P.temp+geom_vline(xintercept=temp.tick,linetype=3,col="grey50")
   }
                                 
   if(opt.SIM.quantile.CI.area){

      temp.simCI<-findQuantileCI(sim.data,plot.data$X,
                                 time.bin,q=q.list,alpha=alpha,N.sim)
      if(!CIarea.box){
         NN<-nrow(temp.simCI$TIME.bin.summary)
         test.LU<-c(temp.simCI$TIME.bin.summary[1,4],
                    temp.simCI$TIME.bin.summary[,6],         
        # test.LU<-c(temp.simCI$TIME.bin.summary[1,2],
        #            temp.simCI$TIME.bin.summary[,2],
                   #apply(temp.simCI$TIME.bin.summary[,4:5],1,mean),
                   temp.simCI$TIME.bin.summary[NN,5])
         test.data.tot<-temp.simCI$Q.CI
         n.temp<-length(test.LU)
         X<-c(test.LU,test.LU[n.temp:1])
         test.data<-test.data.tot[[1]]
         NN<-nrow(test.data)
         Y<-c(test.data[1,2],test.data[,2],test.data[NN,2],
                  test.data[NN,4],test.data[NN:1,4],test.data[1,4])
      } else{
         test.LU<-temp.simCI$TIME.bin.summary[,4:5]
         test.data.tot<-temp.simCI$Q.CI
         X.temp<-c(test.LU[,1],test.LU[nrow(test.LU),2])
         n.temp<-nrow(test.LU)
         X<-c(test.LU[1,1],rep(test.LU[2:n.temp,1],each=2),test.LU[n.temp,2])
         X<-c(X,X[length(X):1])
         test.data<-test.data.tot[[1]]
         Y<-c(rep(test.data[,2],each=2),rep(test.data[(n.temp:1),4],each=2))
      }
      if(Y0.adjust) Y[Y<0]<-0
      SIM.CIarea.1<-data.frame(X=X,Y=Y,ID=1)

      P.temp<-P.temp+geom_polygon(data= SIM.CIarea.1,
                                  aes(x=X,y=Y,group=ID,fill=ID),
                                  fill="lightblue",colour="lightblue")
                                  #fill="gray80",colour="gray80")
      test.data<-test.data.tot[[3]]
      if(!CIarea.box){
         NN<-nrow(test.data)
         Y<-c(test.data[1,2],test.data[,2],test.data[NN,2],
              test.data[NN,4],test.data[NN:1,4],test.data[1,4])
      } else{
         Y<-c(rep(test.data[,2],each=2),
              rep(test.data[(n.temp:1),4],each=2))
      }
      if(Y0.adjust) Y[Y<0]<-0      
      SIM.CIarea.3<-data.frame(X=X,Y=Y,ID=1)
      P.temp<-P.temp+geom_polygon(data= SIM.CIarea.3,
                                  aes(x=X,y=Y,group=ID,fill=ID),
                                  fill="lightblue",colour="lightblue")
                                  #fill="gray80",colour="gray80")
      test.data<-test.data.tot[[2]]
      if(!CIarea.box){
         NN<-nrow(test.data)      
         Y<-c(test.data[1,2],test.data[,2],test.data[NN,2],
              test.data[NN,4],test.data[NN:1,4],test.data[1,4])
      } else{
         Y<-c(rep(test.data[,2],each=2),
              rep(test.data[(n.temp:1),4],each=2))
      }
      if(Y0.adjust) Y[Y<0]<-0      
      SIM.CIarea.2<-data.frame(X=X,Y=Y,ID=1)
      P.temp<-P.temp+geom_polygon(data=SIM.CIarea.2,
                                  aes(x=X,y=Y,group=ID,fill=ID),
                                  fill="pink",colour="pink")  
                                  #fill="gray50",colour="gray50")  
   } 
   if(opt.DV.point){
      P.temp<-P.temp+geom_point(color="grey30",size=2,alpha=0.5)
      DV.point<-data.frame(X=orig.data[,X.name],Y=orig.data[,Y.name])
   } 

   if(opt.DV.quantile.line){
      temp<-findQuantile(plot.data$Y,plot.data$X,time.bin,q=q.list)
      if(Y0.adjust) 
        temp[,6:8]<-apply(temp[,6:8],2,function(x) {x[x<0]<-0;x})
      DV.quant<-data.frame(X=rep(apply(temp[,4:5],1,mean),length(q.list)),
                           G=factor(rep(paste("Q",round(q.list*100),
                                            "th",sep=""),each=nrow(temp))),
                           Y=unlist(temp[,-(1:(ncol(temp)-length(q.list)))]))
      
         P.temp<-P.temp+geom_line(data = DV.quant[DV.quant$G!="Q50th",],
                                  aes(x=X,y=Y,group=G),
                                  linetype=1,size=1,color="blue")+
                                  #linetype=1,size=1,color="black")+
                        geom_line(data = DV.quant[DV.quant$G=="Q50th",],
                                  aes(x=X,y=Y,group=G),
                                  linetype=1,size=1,color="red")
                                  #linetype=1,size=1,color="black")
         if(orig.flag){
            P.temp<-P.temp+geom_point(data = DV.quant[DV.quant$G!="Q50th",],
                                 aes(x=X,y=Y,group=G),
                                 size=4,color="blue")+
          #linetype=1,size=1,color="black")+
                      geom_point(data = DV.quant[DV.quant$G=="Q50th",],
                                 aes(x=X,y=Y,group=G),
                                 size=4,color="red")
         }
   } 
 
   if(opt.SIM.quantile.line){
      temp.sim.Q<-findQuantile(sim.data,plot.data$X,time.bin,q=q.list)
      if(Y0.adjust) 
        temp.sim.Q[,6:8]<-apply(temp.sim.Q[,6:8],2,function(x) {x[x<0]<-0;x})      
      SIM.quant<-data.frame(X=rep(temp.sim.Q$med.COV,length(q)),
                            G=factor(rep(paste("Q",round(q.list*100),
                                       "th",sep=""),each=nrow(temp.sim.Q))),
                            Y=unlist(temp.sim.Q[,-(1:(ncol(temp.sim.Q)-
                                                        length(q.list)))]))

      P.temp<-P.temp+geom_line(data = SIM.quant[SIM.quant$G!="Q50th",],
                              aes(x=X,y=Y,group=G),linetype=2,
                              size=1,color="blue")+
                    geom_line(data = SIM.quant[SIM.quant$G=="Q50th",],
                              aes(x=X,y=Y,group=G),linetype=2,
                              size=1,color="red")
   }
   if(plot.flag){
      P.temp+theme_bw()+theme(panel.grid.major=element_line(colour = "white"))+
                        theme(panel.grid.minor=element_line(colour="white"))
   } else{
      return(list(SIM.CIarea.1=SIM.CIarea.1,SIM.CIarea.2=SIM.CIarea.2, 
                SIM.CIarea.3=SIM.CIarea.3,DV.point=DV.point,DV.quant=DV.quant,
                SIM.quant=SIM.quant))
  }
}