simdata: Reference data.

In bamlss: Bayesian Additive Models for Location, Scale, and Shape (and Beyond)

Reference data.

Description

Simulated data to test the implementation of the bamlss families.

Usage

data("simdata")

Format

An object of class list of length 3.

See Also

mvnchol_bamlss

Examples

## Not run: ## Reproducing code.
set.seed(111)
n <- 2000

## build orthogonal rotation matrix
thetax <- pi/4
thetay <- pi/4
thetaz <- pi/4
Rx <- matrix( c(1,0,0, 0,cos(thetax),sin(thetax), 0,-sin(thetax),cos(thetax) ), 3, 3 )
Ry <- matrix( c(cos(thetay),0,-sin(thetay), 0,1,0,  sin(thetay),0,cos(thetay) ), 3, 3 )
Rz <- matrix( c(cos(thetaz),sin(thetaz),0, -sin(thetaz),cos(thetaz),0, 0,0,1 ), 3, 3 )
R <- Rx %*% Ry %*% Rz

## non-linear functions
f1 <- function(x) (sin(pi * x))^2
f2 <- function(x) (cos(pi * x))^2

## random derivitates  
x <- runif(n)

## eigenvalues
val1 <- f1(x) 
val2 <- f2(x)
val3 <- rep(0, n)

## initialize vectors for parameter lists
p12 <- NULL
p13 <- NULL
p23 <- NULL
sig <- matrix(0, n, 3)

lamdiag <- matrix(0, n, 3)
lambda <- matrix(0, n, 3)

y <- matrix(0, n, 3)
log_dens_ref <- rep(0, n)

tau <- .1  ## offset on diagonal
l <- 0     ## count occasions with invertible cv 
dens1 <- NULL
for ( ii in seq(n) ) {
    mu <- rep(0, 3)
  
    val <- diag( c(val1[ii], val2[ii], val3[ii]) ) + diag(tau, 3)
    ## compute covariance matrix from rotation matrix and eigenvalues
    cv <- R %*% val %*% t(R)
  
    ## compute parameters for parameter list
    sig[ii,] <- sqrt(diag(cv))
    p12[ii] <- cv[1,2]
    p13[ii] <- cv[1,3]
    p23[ii] <- cv[2,3]
  
    ## compute paramters for Cholesky family
    chol_cv <- solve(chol(cv))  # lambdas come from L^-1 not L
    lamdiag[ii,] <- diag(chol_cv)
    lambda[ii,] <- chol_cv[upper.tri(chol_cv)]
  
    ## Check if cv is invertible 
    if ( !is.matrix(try(chol(cv))) ) l <- l + 1
  
    y[ii,] <- mvtnorm::rmvnorm(1, mu, cv)
  
    log_dens_ref[ii] <- mvtnorm::dmvnorm(y[ii,], mu, cv, log = TRUE)
}
print(l)

## Data
d <- as.data.frame(y)
names(d) <- paste0("y", 1:3)
d$x <- x

## make parameter list for mvn chol family
par <- list()
par[["mu1"]] <- rep(0,n)
par[["mu2"]] <- rep(0,n)
par[["mu3"]] <- rep(0,n)
par[["lamdiag1"]] <- lamdiag[,1]
par[["lamdiag2"]] <- lamdiag[,2]
par[["lamdiag3"]] <- lamdiag[,3]
par[["lambda12"]] <- lambda[,1]
par[["lambda13"]] <- lambda[,2]
par[["lambda23"]] <- lambda[,3]

simdata <- list(
    d   = d,
    par = par,
    y   = y
)

## save(simdata, file = "simdata.rda")
## End of simulation

## End(Not run)

bamlss documentation built on Oct. 11, 2024, 5:07 p.m.