whatmv: Calculate weights for deconvolution of bivariate KDE

In DeconWK: Deconvolution by Weighted Kernels

Description

Routine to calculate the weights for deconvolution via weighted kernel density estimates for the bivariate case.

Usage

w.hat.mv(y, Sigma, H, gamma, ...)

Arguments

y	a matrix with two columns containing the observed, contaminated data.
Sigma	the variance-covariance matrix of the contaminating (normal) distribution.
H	the matrix of smoothing parameters to be used for the weighted bivariate kernel density estimate; if missing the bandwidth returned by Hpi(y,) will be used.
gamma	the regularisation parameter to be used; either a scalar or a vector of values from which a suitable value is selected via K-fold cross-validation.
...	optional parameters passed to the cross-validation routine cv.score.mv if K-fold cross-validation is used.

Value

A vector containing the weights; if gamma is chosen by cross-validation, the selected value is returned as an attribute.

Author(s)

Martin L Hazelton [e-mail]

Berwin A Turlach [e-mail]

References

Hazelton, M.L. and Turlach, B.A. (2009). Nonparametric density deconvolution by weighted kernel estimators, Statistics and Computing 19(3): 217–228. http://dx.doi.org/10.1007/s11222-008-9086-7.

See Also

wkde.2d, wkde.contour, Hpi

Examples

  library(ks)
  Age <- framingham[,2]
  Age.lim.2 <- 56
  SBP1.A <- framingham[Age>=Age.lim.2,3] # SBP, measure 1, Exam 2 
  SBP2.A <- framingham[Age>=Age.lim.2,4] # SBP, measure 2, Exam 2 
  SBP1.B <- framingham[Age>=Age.lim.2,5] # SBP, measure 1, Exam 3 
  SBP2.B <- framingham[Age>=Age.lim.2,6] # SBP, measure 2, Exam 3 
  sigma.fram.A <- sd(SBP1.A-SBP2.A)
  sigma.fram.B <- sd(SBP1.B-SBP2.B)
  Sigma.fram <- diag(c(sigma.fram.A,sigma.fram.B))^2
  SBP.A <- SBP1.A
  SBP.B <- SBP1.B
  SBP.bi <- cbind(SBP.A,SBP.B)
  H.fram <- Hpi(SBP.bi)

  w <-  w.hat.mv(SBP.bi, Sigma.fram, H.fram, gamma = 0.4)
  plot(SBP.bi, cex=w)  

DeconWK documentation built on May 2, 2019, 6:08 p.m.