Last chance! 50% off unlimited learning
Sale ends in
Calculates the proportion of particles which use each input to make a tree split and the proportion of all splits in trees of each particle that correspond to each input variable; also provides tree height and leaf size summary information
# S3 method for dynaTree
varpropuse(object)
# S3 method for dynaTree
varproptotal(object)
# S3 method for dynaTree
treestats(object)For varprop*, a
vector of proportions of length ncol(object$X)) is returned;
for treestats a 1-row, 4-column data.frame is
returned
a "dynaTree"-class object built by dynaTree
Robert B. Gramacy rbg@vt.edu,
Matt Taddy and Christoforos Anagnostopoulos
varpropuse gives the proportion of times a particle
uses each input variable in a tree split; varproptotal gives
the proportion of total uses by the tree in each particle (i.e.,
averaged over the total number of splits used in the tree).
Usually, varpropuse returns a vector of (nearly) all ones
unless there are variables which are not useful in predicting
the response. Using model = "linear" is not recommended
for this sort of variable selection.
treestats returns the average tree height, and the average
leaf size, both active and retired
Gramacy, R.B., Taddy, M.A., and S. Wild (2011). “Variable Selection and Sensitivity Analysis via Dynamic Trees with an Application to Computer Code Performance Tuning” arXiv:1108.4739
dynaTree, sens.dynaTree,
relevance.dynaTree
## ffit a dynaTree model to the Ozone data
X <- airquality[,2:4]
y <- airquality$Ozone
na <- apply(is.na(X), 1, any) | is.na(y)
out <- dynaTree(X=X[!na,], y=y[!na])
## obtain variable usage proportions
varpropuse(out)
varproptotal(out)
## gather relevance statistics which are more meaningful
out <- relevance(out)
boxplot(out$relevance)
abline(h=0, col=2, lty=2)
## obtain tree statistics
treestats(out)
## clean up
deletecloud(out)
Run the code above in your browser using DataLab