eval.stage {delt} | R Documentation |
Returns a stagewise minimization estimate. A stagewise minimization estimator is a convex combination of greedy histograms. The convex combination is constructed by a stagewise minimization of an empirical risk functional.
eval.stage(dendat, leaf, M, pis = NULL, mcn = dim(dendat)[1], minobs = NULL, seedi = 1, method = "projec", bound = 0)
dendat |
n*d data matrix |
leaf |
the (maximal) number of rectangles in the partition of the greedy histograms |
M |
the number of histograms in the convex combination |
pis |
the vector of weights of the convex combination |
mcn |
the size of the Monte Carlo sample used in the numerical integration in calculating the empirical risk functional |
minobs |
non-negative integer; splitting of a bin of a greedy histogram will be continued if the bin containes "minobs" or more observations |
seedi |
the seed for the generation of the Monte Carlo sample |
method |
"loglik" or "projec"; the empirical risk is either the log-likelihood or the L2 empirical risk |
bound |
internal |
An evaluation tree
Jussi Klemela
dendat<-sim.data(n=200,seed=5,type="mulmodII") leaf<-13 # the number of leafs of the greedy histograms M<-5 # the number of greedy histograms pcf<-eval.stage(dendat,leaf=leaf,M=M) dp<-draw.pcf(pcf,pnum=c(120,120)) persp(dp$x,dp$y,dp$z,ticktype="detailed",phi=25,theta=-120)