stabilitySelection: Selection of the regularization parameter by StARS (Liu et...

In RNAseqNet: Log-Linear Poisson Graphical Model with Hot-Deck Multiple Imputation

Description

stabilitySelection implements the regularization parameter selection of (Liu et al., 2010) called 'Stability Approach to Regularization Selection' (StARS).

Usage

stabilitySelection(counts, lambdas = NULL, B = 20)

Arguments

counts	a n x p matrix of RNA-seq expression (numeric matrix or data frame)
lambdas	a sequence of decreasing positive numbers to control the regularization (numeric vector). Default to NULL
B	number of iterations for stability selection. Default to 20

Details

When input lambdas are null the default sequence of glmnet (see GLMnetwork for details).

Value

S3 object of class stabilitySelection : a list consisting of

• lambdas numeric regularization parameters used for regularization path

• B number of iterations for stability selection

• best index of the regularization parameter selected by StARS in lambdas

• variabilities numeric vector having same length than lambdas and providing the variability value as defined by StARS along the path

Author(s)

Alyssa Imbert, [e-mail] Nathalie Vialaneix, [e-mail]

References

Liu, H., Roeber, K. and Wasserman, L. (2010) Stability approach to regularization selection (StARS) for high dimensional graphical models. In Proceedings of Neural Information Processing Systems (NIPS 2010), 23, 1432-1440, Vancouver, Canada.

See Also

GLMnetwork

Examples

data(lung)
lambdas <- 4 * 10^(seq(0, -2, length = 5))
stability_lung <- stabilitySelection(lung, lambdas = lambdas, B = 4)
## Not run: plot(stability_lung)

RNAseqNet documentation built on July 2, 2020, 4:15 a.m.