DICTAUG: Performs a grid search over the marginal posterior...

In AndrewGChapple/SCRSELECT: Performs Bayesian Variable Selection on the Covariates in a Semi-Competing Risks Model

Description

Performs a grid search over the marginal posterior probabilities of inclusion and returns a list of DIC values corresponding to each grid point. This is used in the ReturnModel function.

Usage

DICTAUG(PCT1, PCT2, PCT3, COV, Y1, Y2, I1, I2, s1, lam1, s2, lam2, s3, lam3,
  gam, c, B, inc)

Arguments

PCT1	Vector Containing posterior probabilities of inclusion for the hazard of a non-terminal event. This must be of length ncol(COV)-inc.
PCT2	Vector Containing posterior probabilities of inclusion for the hazard of death without a non-terminal event. This must be of length ncol(COV)-inc.
PCT3	Vector Containing posterior probabilities of inclusion for the hazard of death after a non-terminal event. This must be of length ncol(COV)-inc.
COV	Matrix of Patient Covariates. The last inc will be left out of variable selection.
Y1	Vector Containing non-terminal event times (or censoring time due to death/censoring).
Y2	Vector Containing Terminal Event times (or censoring).
I1	Vector Containing non-terminal event indicators (1 if non-terminal event for a patient, 0 otherwise).
I2	Vector Containing Terminal event indicators (1 if a patients experiences a non-ternminal event, 0 if censored).
s1	Vector containing the posterior locations of the split points in the hazard of a non-terminal event.
lam1	Vector containing the posterior log hazard heights on the split point intervals in the hazard of a non-terminal event.
s2	Vector containing the posterior locations of the split points in the hazard of death without a non-terminal event.
lam2	Vector containing the posterior log hazard heights on the split point intervals in the hazard of death without a non-terminal event.
s3	Vector containing the posterior locations of the split points in the hazard of death after a non-terminal event.
lam3	Vector containing the posterior log hazard heights on the split point intervals in the hazard of death after a non-terminal event.
gam	Vector of length n containing the posterior mean frailties of the patients.
c	Hyperparameter involved in the sampling of hazard coefficients. This should be the same value that controls the degree of sparsity achieved by the SVSS.
B	Number of iterations
inc	Number of variables left out of selection

Value

Returns a list of size 18 containing 18x18 matrices of DIC values and skipped entries.

@references [1] Lee, K. H., Haneuse, S., Schrag, D. and Dominici, F. (2015), Bayesian semi-parametric analysis of semi-competing risks data: investigating hospital readmission after a pancreatic cancer diagnosis. Journal of the Royal Statistical Society: Series C (Applied Statistics), 64: 253-273. doi: 10.1111/rssc.12078 [2] Chapple, A.C., Vannucci, M., Thall, P.F., Lin, S.(2017), Bayesian Variable selection for a semi-competing risks model with three hazard functions. Journal of Computational Statistics & Data Analysis, Volume 112, August 2017, Pages 170-185 [3] https://adventuresinstatistics.wordpress.com/2017/04/10/package-scrselect-using-returnmodel/

Examples

####Randomly Generate Semicompeting Risks Data
####Generates random patient time, indicator and covariates.
set.seed(1)
n=100
Y1=runif(n,0,100)
I1=rbinom(n,1,.5)
Y2=Y1
I2=I1
for(i in 1:n){if(I1[i]==0){Y2[i]=Y1[i]}else{Y2[i]=Y1[i]+runif(1,0,100)}}
I2=rbinom(n,1,.5)
library(mvtnorm)
X=rmvnorm(n,rep(0,7),diag(7))
###Read in Posterior mean quantities from SCRSELECTRETURN
PCT1=c(.2,.4,.7,.8,.5)
PCT2=c(.02,.06,.1,.5,.7)
PCT3=c(.85,.87,.3,.45,.51)
gam=rgamma(n,1,1)
s1=c(0,3,5,max(Y1[I1==1]))
lam1=c(-1,-3,0)
s2=c(0,1,max(Y2[I1==0]))
lam2=c(0,-2)
s3=c(0,max(Y2[I1==1]))
lam3=-2
####Read in Hyperparameters
c=5
###Number of iterations and output location
B=4
###Number of variables to exclude from selection and burnin percent
inc=2
DICTAUG(PCT1,PCT2,PCT3,X,Y1,Y2,I1,I2,s1,lam1,s2,lam2,s3,lam3,gam,c,B,inc)

AndrewGChapple/SCRSELECT documentation built on May 28, 2019, 11:04 a.m.