UMR (version 1.1.0)

UMRgrad_generic: Gradient of least-squares Shuffled Isotonic Regression criterion

Description

Gradient of least-squares Shuffled Isotonic Regression criterion

Usage

UMRgrad_generic(
  yy,
  ww_y = rep(1/length(yy), length(yy)),
  mm,
  ww_m = rep(1/length(mm), length(mm)),
  AAfunc,
  BBfunc
)
grad_SIR_generic(
  yy,
  mm,
  counts = rep(1, length(mm)),
  AAfunc,
  BBfunc,
  rescale = FALSE
)

Arguments

yy

Y (response) observation vector (numeric)

ww_y

Weight vector for yy.

mm

Current (unsorted) estimate/iterate at which to compute gradient. (Length equals length of yy).

ww_m

Weight vector for mm.

AAfunc

This is the function "A" defined in the gradient calculations in the paper (Balabdaoui, Doss, Durot (2020+).

BBfunc

This is the function "B" defined in the gradient calculations in the paper (Balabdaoui, Doss, Durot (2020+).

@details Returns gradient as a column matrix. See calculations in the paper.

@examples #### See help for gradDesc_PC, gradDesc, or grad_helpers

counts

If the function that mm represents is piecewise constant, then mm may be passed in as only the unique entries. In that case counts contains the number of times each element of mm is repeated. Thus length(counts)==length(mm). (Default for counts is thus a vector of all 1's.)

rescale

Boolean: if False then the final return value is the