R/invM2.n.R

In RobustAFT: Truncated Maximum Likelihood Fit and Robust Accelerated Failure Time Regression for Gaussian and Log-Weibull Case

"invM2.n" <-
function(u,theta,sigma,rs,wi,XtX,xbar,estim=c("SA","TMLA","SI","TMLI")) {
n <- length(rs); p <- ncol(XtX); xk <- 1.5477
if (estim=="SA")  {zpsp <- psp.weight(rs,ips=2,xk=xk)
                   a1c  <- sum(zpsp)/(n*sigma)
                   b1c  <- sum(zpsp*rs)/(n*sigma)
                   zpsi <- psi.weight(rs,ips=2,xk=xk)
                   a2c  <- sum(zpsi)/(n*sigma)
                   b2c  <- sum(zpsi*rs)/(n*sigma)}
if (estim=="TMLA"){a1c  <- sum(wi)/(n*sigma)
                   b1c  <- sum(rs*wi)/(n*sigma)
                   a2c  <- 2*b1c
                   b2c  <- 2*sum(rs^2*wi)/(n*sigma)}
if (estim=="SI")  {a1c   <- integrate(Pspphi.n, lower=-xk,upper=xk)$value/sigma
                   b2c   <- integrate(Psizphi.n,lower=-xk,upper=xk)$value/sigma
                   a2c   <- 0; b1c <- 0}
if (estim=="TMLI"){a1c   <- (2*pnorm(u)-1)/sigma
                   b2c   <- 4*(-u*dnorm(u)+pnorm(u)-0.5)/sigma
                   a2c   <- 0; b1c <- 0}
A1 <- a1c*XtX;    b1 <- b1c*xbar
a2 <- a2c*xbar;   b2 <- b2c
M  <- matrix(0,ncol=p+1,nrow=p+1)
M[1:p,1:p] <- A1; M[1:p,p+1] <- as.matrix(b1)
M[p+1,1:p] <- a2; M[p+1,p+1] <- b2
Minv <- solve(M)
list(Minv=Minv)}