inst/md/censored.md

In gdlc/BGLR-R: Bayesian Generalized Linear Regression

Censored Regression

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

Case | $a_i$ | $y_i$ | $b_i$ -------------------|-----|------|---- Not-censored | NA | $y_i$ | NA Right-Censored | $a_i$ | NA | Inf Left-Censored |-Inf | NA | $b_i$ Interval-Censored | $a_i$ | NA | $b_i$

 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)

Back to examples

gdlc/BGLR-R documentation built on April 23, 2024, 11:01 p.m.