View source: R/Transformations.R
| fcenLRu | R Documentation |
Maps a P-density from the weighted Bayes space B^2(P)
to the unweighted space L^2(\lambda) by multiplying the
P-weighted clr transform (fcenLRp) by \sqrt{p},
\mathrm{fcenLR}_u(f)(z) = \sqrt{p(z)}\;\mathrm{fcenLR}_P(f)(z).
This moves the weighting from the measure into the function, so that methods
for unweighted L^2 data (e.g. functional PCA) can be applied.
fcenLRu(z, z_step, density, p)
z |
numeric vector, equidistant grid of points on the domain. |
z_step |
spacing of the grid |
density |
numeric vector of the same length as |
p |
numeric vector of the same length as |
Integrals are approximated by trapzc, so z must be
an equidistant grid with spacing z_step. All values of density
must be strictly positive. Back-transforming the result with
fcenLRinv gives the representation of the density in the
unweighted Bayes space B^2(\lambda).
A numeric vector of the same length as density: the unweighted
clr transform evaluated on the grid (an element of L^2(\lambda)).
R. Talska talskarenata@seznam.cz, A. Menafoglio, K. Hron karel.hron@upol.cz, J. J. Egozcue, J. Palarea-Albaladejo
Talska, R., Menafoglio, A., Hron, K., Egozcue, J. J., Palarea-Albaladejo, J. (2020). Weighting the domain of probability densities in functional data analysis. Stat. \Sexpr[results=rd]{tools:::Rd_expr_doi("10.1002/sta4.283")}
fcenLR, fcenLRp, fcenLRinv,
trapzc
# Common example for all transformations - fcenLR, fcenLRp, fcenLRu
# Example (log normal distribution under the reference P)
t = seq(1,10, length = 1000)
t_step = diff(t[1:2])
# Log normal density w.r.t. Lebesgue reference measure in B2(lambda)
f = dlnorm(t, meanlog = 1.5, sdlog = 0.5)
# Log normal density w.r.t. Lebesgue reference measure in L2(lambda)
f.fcenLR = fcenLR(t,t_step,f)
# New reference given by exponential density
p = dexp(t,0.25)/trapzc(t_step,dexp(t,0.25))
# Plot of log normal density w.r.t. Lebesgue reference measure
# in B2(lambda) together with the new reference density p
matplot(t,f,type="l",las=1, ylab="density",cex.lab=1.2,cex.axis=1.2,
col="black",lwd=2,ylim=c(0,0.3),xlab="t")
matlines(t,p,col="blue")
text(2,0.25,"p",col="blue")
text(4,0.22,"f",col="black")
# Log-normal density w.r.t. exponential distribution in B2(P)
# (unit-integral representation)
fp = (f/p)/trapzc(t_step,f/p)
# Log-normal density w.r.t. exponential distribution in L2(P)
fp.fcenLRp = fcenLRp(t,t_step,fp,p)
# Log-normal density w.r.t. exponential distribution in L2(lambda)
fp.fcenLRu = fcenLRu(t,t_step,fp,p)
# Log-normal density w.r.t. exponential distribution in B2(lambda)
fp.u = fcenLRinv(t,t_step,fp.fcenLRu)
# Plot
layout(rbind(c(1,2,3,4),c(7,8,5,6)))
par(cex=1.1)
plot(t, f.fcenLR, type='l', ylab=expression(fcenLR[lambda](f)),
xlab='t',las=1,ylim=c(-3,3),
main=expression(bold(atop(paste('(a) Representation of f in ', L^2, (lambda)),'[not weighted]'))))
abline(h=0,col="red")
plot(t, f, type='l', ylab=expression(f[lambda]),
xlab='t',las=1,ylim=c(0,0.4),
main=expression(bold(atop(paste('(b) Density f in ', B^2, (lambda)),'[not weighted]'))))
plot(t, fp, type='l', ylab=expression(f[P]), xlab='t',
las=1,ylim=c(0,0.4),
main=expression(bold(atop(paste('(c) Density f in ', B^2, (P)),'[weighted with P]'))))
plot(t, fp.fcenLRp, type='l', ylab=expression(fcenLR[P](f[P])),
xlab='t',las=1,ylim=c(-3,3),
main=expression(bold(atop(paste('(d) Representation of f in ', L^2, (P)),'[weighted with P]'))))
abline(h=0,col="red")
plot(t, fp.u, type='l', ylab=expression(paste(omega^(-1),(f[P]))),
xlab='t',las=1,ylim=c(0,0.4),
main=expression(bold(atop(paste('(e) Representation of f in ', B^2, (lambda)),'[unweighted]'))))
plot(t, fp.fcenLRu, type='l', ylab=expression(paste(fcenLR[u](f[P]))),
xlab='t',las=1,ylim=c(-3,3),
main=expression(bold(atop(paste('(f) Representation of f in ', L^2, (lambda)),'[unweighted]'))))
abline(h=0,col="red")
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.