censored.md
In BGLR: Bayesian Generalized Linear Regression

BGLR supports right, left and interval censoring. For censored outcomes the response is represented with a triplet ai,yi,bi, where ai and bi are the lower and upper bounds for the phenotype (yi). The following table describes the configuration of the tripled for un-censored, right, left and interval censored data

Case | ai | yi | bi -------------------|-----|------|---- Not-censored | NA | yi | NA Right-Censored | ai | NA | Inf Left-Censored |-Inf | NA | bi Interval-Censored | ai | NA | bi

 library(BGLR); data(wheat)
 y=wheat.Y[,1]

 # Introducing censoring (for simplicity done at fixed points -1 and 1)
 yNA=y
 a=rep(NA,599)
 b=rep(NA,599)
 type=sample(1:4,prob=c(2/3,1/9,1/9,1/9),size=599,replace=T)
# left-censored
 a[type==2]=-Inf
 b[type==2]=1
 yNA[type==2]=NA


# right-censored
 a[type==2]=-1
 b[type==2]=Inf
 yNA[type==2]=NA

# right-censored
 a[type==2]=-1
 b[type==2]=1
 yNA[type==2]=NA 

 fm=BGLR(y=yNA,a=a,b=b)
 plot(y,fm$yHat)