dbivr: Bivariate kernel density estimation for rounded data

In Kernelheaping: Kernel Density Estimation for Heaped and Rounded Data

Description

Bivariate kernel density estimation for rounded data

Usage

dbivr(
  xrounded,
  roundvalue,
  burnin = 2,
  samples = 5,
  adaptive = FALSE,
  gridsize = 200
)

Arguments

xrounded	rounded values from which to estimate bivariate density, matrix with 2 columns (x,y)
roundvalue	rounding value (side length of square in that the true value lies around the rounded one)
burnin	burn-in sample size
samples	sampling iteration size
adaptive	set to TRUE for adaptive bandwidth
gridsize	number of evaluation grid points

Value

The function returns a list object with the following objects (besides all input objects):

Mestimates	kde object containing the corrected density estimate
gridx	Vector Grid on which density is evaluated (x)
gridy	Vector Grid on which density is evaluated (y)
resultDensity	Array with Estimated Density for each iteration
resultX	Matrix of true latent values X estimates
delaigle	Matrix of Delaigle estimator estimates

Examples

# Create Mu and Sigma  -----------------------------------------------------------
mu1 <- c(0, 0)
mu2 <- c(5, 3)
mu3 <- c(-4, 1)
Sigma1 <- matrix(c(4, 3, 3, 4), 2, 2)
Sigma2 <- matrix(c(3, 0.5, 0.5, 1), 2, 2)
Sigma3 <- matrix(c(5, 4, 4, 6), 2, 2)
# Mixed Normal Distribution -------------------------------------------------------
mus <- rbind(mu1, mu2, mu3)
Sigmas <- rbind(Sigma1, Sigma2, Sigma3)
props <- c(1/3, 1/3, 1/3)
## Not run: xtrue=rmvnorm.mixt(n=1000, mus=mus, Sigmas=Sigmas, props=props)
roundvalue=2
xrounded=plyr::round_any(xtrue,roundvalue)
est <- dbivr(xrounded,roundvalue=roundvalue,burnin=5,samples=10)

#Plot corrected and Naive distribution
plot(est,trueX=xtrue)
#for comparison: plot true density
 dens=dmvnorm.mixt(x=expand.grid(est$Mestimates$eval.points[[1]],est$Mestimates$eval.points[[2]]),
  mus=mus, Sigmas=Sigmas, props=props)
 dens=matrix(dens,nrow=length(est$gridx),ncol=length(est$gridy))
 contour(dens,x=est$Mestimates$eval.points[[1]],y=est$Mestimates$eval.points[[2]],
    xlim=c(min(est$gridx),max(est$gridx)),ylim=c(min(est$gridy),max(est$gridy)),main="True Density")
## End(Not run)

Kernelheaping documentation built on Jan. 27, 2022, 1:09 a.m.