qic: Calculate information criterion for penalized quantile...

In bssherwood/rqpen: Penalized Quantile Regression

Calculate information criterion for penalized quantile regression models. Currently not exported.

Description

Calculate information criterion for penalized quantile regression models. Currently not exported.

Usage

qic(model, n, method = c("BIC", "AIC", "PBIC"))

Arguments

model	model from a rq.pen.seq() object
n	Sample size
method	Choice of BIC, AIC or PBIC, a large p BIC.

Value

Let \hat{\beta} be the coefficient vectors for the estimated model. Function returns the value

\log(\sum_{i=1}^n w_i \rho_\tau(y_i-x_i^\top\hat{\beta})) + d*b/(2n),

where d is the number of nonzero coefficients and b depends on the method used. For AIC b=2, for BIC b=log(n) and for PBIC d=log(n)*log(p) where p is the dimension of \hat{\beta}. The values of w_i default to one and are set using weights when fitting the models. Returns this value for each coefficient vector in the model, so one for every value of \lambda.

Author(s)

Ben Sherwood, [e-mail]

References

\insertRef

qrbicrqPen

Examples

## Not run: 
set.seed(1)
x <- matrix(runif(800),ncol=8)
y <- 1 + x[,1] + x[,8] + (1+.5*x[,3])*rnorm(100)
m1 <- rq.pen(x,y,tau=c(.25,.75))
# returns the IC values for tau=.25
qic(m1$models[[1]],m1$n) 
# returns the IC values for tau=.75
qic(m1$models[[2]],m1$n)

## End(Not run) 

bssherwood/rqpen documentation built on April 23, 2024, 9:50 a.m.