tests/expectations.R

In robustlmm: Robust Linear Mixed Effects Models

## Test disable
quit()

## test .calcE.D.re and .calcEchi
require(robustlmm)

.calcE.D.re <- robustlmm:::.calcE.D.re
cPsi <- robustlmm:::cPsi
.d <- robustlmm:::.d
.d2 <- robustlmm:::.d2

## simple tests for q = 1:
test1 <- function(rho) {
    stopifnot(all.equal(.calcE.D.re(1, rho), rho@EDpsi()))
}

test1(cPsi)
test1(smoothPsi)

## check very large q
stopifnot(all.equal(.calcE.D.re(10000, huberPsiRcpp), 0),
          all.equal(.calcE.D.re(10000, smoothPsi), 0))

## tests for q > 1:
test2 <- function(q, rho, npts = 1000000) {
    cat("\nq", q, "\n")
    ## test .calcE.D.re
    integrand <- function(u2, v2, s2) {
        dk <- .d2(u2 + v2,q)
        (rho@Dpsi(dk)/dk - rho@psi(dk)/dk^2)*2*u2 + rho@psi(.d2(s2,q))/.d2(s2,q)
    }
    u2 <- rnorm(npts)^2
    v2 <- rchisq(npts, q-1)
    s2 <- rchisq(npts, q)
    value <- mean(integrand(u2, v2, s2))
    cat("\nImportance Sampling:", value, "\n")
    cat(".calcE.D.re:        ", .calcE.D.re(q, rho), "\n")
    #print(all.equal(value, .calcE.D.re(q, rho)))
    stopifnot(all.equal(value, .calcE.D.re(q, rho), tolerance = 1e-2))
}

for (q in c(2,4)) test2(q, cPsi)
for (q in c(2,4,10)) test2(q, smoothPsi)


require(robustlmm)

integrand <- function(a2, b2) {
       d <- sqrt(a2 + b2)
       (rho@Dpsi(d)/d - rho@psi(d)/d^2)/d*a2 + rho@wgt(d)
}

npts <- 10000000
a <- rnorm(npts)
b <- rnorm(npts)
rho <- smoothPsi

value <- mean(integrand(a * a, b * b))

value
robustlmm:::.calcE.D.re(2, smoothPsi)



cat('Time elapsed: ', proc.time(),'\n') # for ``statistical reasons''