cic: Covariance inflation criterion

Description Usage Arguments Value Author(s) References See Also Examples

Description

Computes the covariance inflation criterion (CIC) of Tibshirani and Knight (1999) for submodels of a full linear model.

Usage

1
cic(y, X, nperms = 499, covests = NULL, nullcic = NULL)

Arguments

y

outcome vector

X

model matrix. This should not include an intercept column; such a column is added by the function.

nperms

number of permuted data sets to generate.

covests

sum of the null-hypothesis covariances between the outcomes and the fitted values for the best linear model of each size. If NULL, covariance is estimated from permuted data.

nullcic

CIC for the intercept-only model.

Value

A list with components

leaps

all-subsets regression object (for the unpermuted data) returned by function leaps in package leaps.

covests

sum of the (estimated) null-hypothesis covariances between the outcomes and the fitted values for the best linear model of each size.

enp

effective number of parameters for models of each size, as defined by Tibshirani and Knight (1999).

cic

CIC for each of the models given in the leaps component.

nullcic

CIC for the intercept-only model.

best

vector of logicals indicating which predictors are included in the minimum-CIC model.

Author(s)

Philip Reiss phil.reiss@nyumc.org and Lei Huang huangracer@gmail.com

References

Tibshirani, R., and Knight, K. (1999). The covariance inflation criterion for adaptive model selection. Journal of the Royal Statistical Society, Series B, 61, 529–546.

See Also

leaps (in the package of the same name)

Examples

1
2
3
data(swiss)
cicobj = cic(swiss$Fertility, swiss[ , -1])
cicobj$best

reams documentation built on May 2, 2019, 2:23 p.m.