getCTechSTM1/2u | R Documentation |
compContourM1/2u
The functions getCTechSTM1u
and getCTechSTM2u
set the default list of options CTechST for computing
all the directional (regression) quantiles by means of
compContourM1u
and compContourM2u
,
respectively.
getCTechSTM1u() getCTechSTM2u()
none
Fortunately, the default list of options usually leads to a satisfactory performance in all but very large problems.
Both getCTechSTM1u
and getCTechSTM2u
produce
a list with a few components whose default values are stated
below after the equality sign.
The components OutFilePrefS
and getCharST
are initialized in a method-specific way.
The components CubRegWiseI
, ArchAllFI
,
and SkipRedI
are relevant only if D2SpecI
is zero or if the dimension of directions/responses is
higher than two, i.e., if the breadth-first search algorithm
is used.
Most of the components are generated by both functions.
Nevertheless, the component SkipRedI
is only generated
by getCTechSTM2u
and used by
compContourM2u
.
The output components are as follows:
ReportI |
= 0; if some information (such as the progress
of computation) is displayed on the screen (1)
or not (0).
The display mode may slightly slow down the
computation, especially when the dimension of
responses is higher than two.
On the other hand, it shows the new value of the
quantile level (Tau) (if the input one has been
changed internally), the initial L2-normed
directional vector used ( |
OutSaveI |
= 0; if the detailed output is stored in file(s) into the working directory (1) or not (0). The file output seems necessary only for very large problems if some information about individual cones has to be recorded (such as all the regression quantile hyperplanes used for the regression quantile contour computation). |
D2SpecI |
= 1; this option is relevant only for bivariate directions/responses and determines if the cones are visited counter-clockwise (1) or by means of the breadth-first search algorithm as in the general case (0). The default option (1) leads to a more precise and reliable computation than the other. |
BriefOutputI |
= 1; if the brief (1) or verbose (0)
output is prepared by
|
CubRegWiseI |
= 1; if the directional space is divided into orthants investigated separately (1) or not (0). On the one hand, the default option (1) splits the problem into smaller ones. On the other hand, it also generates some artificial cones with at least one facet in the orthant borders. |
ArchAllFI |
= 1; if all the past cone facet
identifiers (1) or only those from the last
few layers (0) are stored during the computation.
The default option (1) makes the computation
more likely to terminate successfully than
the other. Unfortunately, it is also slower
and more memory demanding. If the dimension
of responses is higher than three, then
|
SkipRedI |
= 0; if the information should be skipped (1)
or stored (0) also from the cones with all
non-artificial facets already known (such cones
are redundant/irrelevant with probability one
if only all the quantile regression
hyperplanes necessary for the quantile contour
computation are required from
|
OutFilePrefS |
= ‘DQOutputM1_’/‘DQOutputM2_’; the prefix of possible output file name(s). |
getCharST |
= getCharSTM1u/getCharSTM2u; the function
computing some overall characteristics that
can be replaced with a user-defined one.
See |
##- a typical use of getCTechSTM1u: ##computing all directional 0.01-quantiles of 49 random points ##(uniformly distributed in the unit cube centered at zero) ##after changing the default settings Tau <- 0.01 XMat <- matrix(1, 49, 1) YMat <- matrix(runif(3*49, -0.5, 0.5), 49, 3) CTechST <- getCTechSTM1u() CTechST$ReportI <- 1 COutST <- compContourM1u(Tau, YMat, XMat, CTechST)
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.