| eval.bagg {delt} | R Documentation |
Returns a bootstrap aggregation of CART-histograms or greedy histograms.
eval.bagg(dendat, B, leaf, minobs = NULL, seed = 1, sample = "bagg", prune = "off", splitscan = 0, seedf = 1, scatter = 0, src = "c", method = "loglik")
dendat |
n*d data matrix |
B |
positive integer; the number of aggregated histograms |
leaf |
the cardinality of the partitions of the aggregated histograms |
minobs |
non-negative integer; a property of aggregated histograms; splitting of a bin will be continued if the bin containes "minobs" or more observations |
seed |
the seed for the random number generation of the random selection of the bootstrap sample |
sample |
"bagg" or "worpl"; the bootstrapping method; "worpl" for the n/2-out-of-n without replacement; "bagg" for n-out-of-n with replacement |
prune |
"on" or "off"; if "on", then CART-histograms will be aggregated; if "off", then greedy histograms will be aggregated |
splitscan |
internal (how many splits will be used for random split selection) |
seedf |
internal (seed for random split selection) |
scatter |
internal (random perturbation of observations) |
src |
internal ("c" or "R" code) |
method |
"loglik" or "projec"; the empirical risk is either the log-likelihood or the L2 empirical risk |
An evaluation tree
Jussi Klemela
lstseq.bagg,
eval.cart,
eval.greedy
dendat<-sim.data(n=600,seed=5,type="mulmodII") leaf<-7 # number of leaves in the histograms seed<-1 # seed for choosing bootstrap samples sample="worpl" # without-replacement bootstrap prune="on" # we use CART-histograms B<-50 # the number of histograms in the average eva<-eval.bagg(dendat,B,leaf,seed=seed,sample=sample,prune=prune) dp<-draw.pcf(eva,pnum=c(60,60)) persp(dp$x,dp$y,dp$z,theta=-20,phi=30)