cpsc_sp {robregcc}

Principal sensitivity component analysis with compositional covariates in sparse setting.

Description

Produce model and its residual estimate based on PCS analysis.

Usage

cpsc_sp(
  X0,
  y0,
  alp = 0.4,
  cfac = 2,
  b1 = 0.25,
  cc1 = 2.937,
  C = NULL,
  we,
  type,
  control = list()
)

Arguments

X0

CLR transformed predictor matrix.

y0

model response vector

alp

(0,0.5) fraction of data sample to be removed to generate subsample

cfac

initial value of shift parameter for weight construction/initialization

b1

tukey bisquare function parameter producing desired breakdown point

cc1

tukey bisquare function parameter producing desired breakdown point

C

sub-compositional matrix

we

penalization index for model parameters beta

type

1/2 for l1 / l2 loss in the model

control

a list of internal parameters controlling the model fitting

Value

betaf

TModel parameter estimate

residuals

residual estimate

References

Mishra, A., Mueller, C.,(2019) Robust regression with compositional covariates. In prepration. arXiv:1909.04990.

Examples

 

library(robregcc)
library(magrittr)

data(simulate_robregcc)
X <- simulate_robregcc$X;
y <- simulate_robregcc$y
C <- simulate_robregcc$C
n <- nrow(X); p <- ncol(X); k <-  nrow(C)

Xt <- cbind(1,X)  # include intercept in predictor
C <- cbind(0,C)    # include intercept in constraint
bw <- c(0,rep(1,p)) # weights not penalize intercept 

example_seed <- 2*p+1               
set.seed(example_seed) 

# Breakdown point for tukey Bisquare loss function 
b1 = 0.5                    # 50% breakdown point
cc1 =  1.567                # corresponding model parameter
b1 = 0.25; cc1 =  2.937   

# Initialization [PSC analysis for compositional data]
control <- robregcc_option(maxiter = 1000,
 tol = 1e-4,lminfac = 1e-7)
fit.init <- cpsc_sp(Xt, y,alp = 0.4, cfac = 2, b1 = b1,
cc1 = cc1,C,bw,1,control)