GetKLEst: Retrieve the surface of Kullback-Leibler (KL) estimators

In sppmix: Modeling Spatial Poisson and Related Point Processes

Description

This function calculates the Kullback-Leibler estimators of the parameters of the components of the mixture intensity, based on a DAMCMC or BDMCMC fit. This is a decision theoretic estimator of the parameters, meaning that, we compute the Posterior Expected Loss (PEL) using the KL loss function and then find the parameter values that minimize the PEL.

For examples see

http://faculty.missouri.edu/~micheasa/sppmix/sppmix_all_examples.html#GetKLEst

Usage

GetKLEst(fit, burnin = floor(fit$L/10), fixLS = FALSE, approx = FALSE,
  segment = 50)

Arguments

fit	Object of class damcmc_res or bdmcmc_res.
burnin	Number of initial realizations to discard. By default, it is 1/10 of the total number of iterations.
fixLS	Logical requesting to check and fix label switching (if present). Default is FALSE.
approx	Logical flag to request use of the identifiability constraint to permute all realizations. Same parameter as in function FixLS_da.
segment	Number of segments to split the posterior realizations into. Each portion of posterior realizations is used to calculate a single Kullback-Leibler realization. The KL estimator is the average of all the KL realizations. Default is 50.

Value

An object of class intensity_surface.

Author(s)

Jiaxun Chen, Sakis Micheas

See Also

rmixsurf, rsppmix, est_mix_damcmc, GetPMEst, GetMAPEst, CompareSurfs

Examples

#generate a surface
truemix_surf <- rmixsurf(m = 3, lambda=100, xlim = c(-3,3), ylim = c(-3,3))
plot(truemix_surf,main="True IPPP intensity surface")
#generate a point pattern
genPPP=rsppmix(intsurf = truemix_surf, truncate = FALSE)
#fit the IPPP model using DAMCMC
fit = est_mix_damcmc(genPPP, m = 3,L=20000)
#get the surfaces of posterior means, MAP and KL estimates
Meansest=GetPMEst(fit)
MAPest=GetMAPEst(fit)
KLest=GetKLEst(fit)
#plot all fitted surfaces
plot(Meansest,main="IPPP intensity surface of posterior means")
plot(MAPest,main="IPPP intensity surface of MAP estimates")
plot(KLest,main="IPPP intensity surface of KL estimates")
#fix labels (if label switching is detected)
KLestLSFixed=GetKLEst(fit,fixLS=TRUE,approx=FALSE)
plot(KLestLSFixed,main="IPPP intensity surface of KL estimates (LS fixed)")
#compare the four estimates against the truth
CompareSurfs(truemix_surf, Meansest, LL = 100, truncate = FALSE)
CompareSurfs(truemix_surf, MAPest, LL = 100, truncate = FALSE)
CompareSurfs(truemix_surf, KLest, LL = 100, truncate = FALSE)
CompareSurfs(truemix_surf, KLestLSFixed, LL = 100, truncate = FALSE)

sppmix documentation built on Jan. 13, 2021, 10:04 p.m.