AIC_jgl: Compute Akaike Information Criterion starting from a jgl...

In GiulioCostantini/JGL2: JGL2: Joint Graphical Lasso, select the lambda parameters using AIC or crossvalidation

Description

Compute Akaike Information Criterion starting from a jgl object, a vector of sample sizes and a list of covariance matrices

Usage

AIC_jgl(jgl, n, S)

Arguments

jgl	The output of JGL
n	a vector of sample sizes.
S	A list of sample covariance matrices

Details

It uses the formula in Danaher et al. (2014)

Value

An integer, the Akaike Information Criterion

Author(s)

Giulio Costantini

References

Danaher, P., Wang, P., and Witten, D. M. (2014). The joint graphical lasso for inverse covariance estimation across multiple classes. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 76(2), 373-397. http://doi.org/10.1111/rssb.12033

See Also

JGL

Examples

## Not run: 
N <- 1000 # sample size
sigma1 <- matrix(c(1, .5, 0, 0,
                   .5, 1, .2, 0,
                   0, .2, 1, 0,
                   0, 0, 0, 1), ncol = 4)

sigma2 <- matrix(c(1, .5, .4, .4,
                   .5, 1, .2, 0,
                   .4, .2, 1, 0,
                   .4, 0, 0, 1), ncol = 4)

dat <- list()
dat[[1]] <- MASS::mvrnorm(n = N, mu = rep(0, ncol(sigma1)), Sigma = sigma1)
dat[[2]] <- MASS::mvrnorm(n = N, mu = rep(0, ncol(sigma2)), Sigma = sigma2)
lapply(dat, function(x) corpcor::cor2pcor(cor(x)))
dat <- data.frame(rbind(dat[[1]], dat[[2]]))
dat$splt <- c(rep(1, N), rep(2, N))
splt <- "splt"
# standardize data data within classes
sp <- split(dat[, !names(dat) == splt], dat[, splt])
sp_sc <- lapply(sp, scale)
dat <- lapply(sp_sc, data.frame)
jgl <- JGL(Y = dat, lambda1 = .1, lambda2 = .1)
S <- lapply(dat, cov)
n <- sapply(dat, nrow)
AIC_jgl(jgl, n, S)

## End(Not run)

GiulioCostantini/JGL2 documentation built on May 6, 2019, 6:29 p.m.