optLambdas_mgcv: Find optimal ridge penalties with maximimum marginal...

In multiridge: Fast Cross-Validation for Multi-Penalty Ridge Regression

Find optimal ridge penalties with maximimum marginal likelihood

Description

Optimizes a marginal likelihood score w.r.t. ridge penalties for multiple data blocks.

Usage

optLambdas_mgcv(penaltiesinit=NULL, XXblocks,Y, pairing=NULL, model=NULL, reltol=1e-4,
  optmethod=ifelse(length(penaltiesinit)==1,"Brent", "Nelder-Mead"),maxItropt=500,
  tracescore=TRUE, fixedpen=NULL, fixedseed =TRUE, sigmasq = 1,
  opt.sigma=ifelse(model=="linear",TRUE, FALSE))

Arguments

penaltiesinit	Numeric vector. Initial values for penaltyparameters. May be obtained from fastCV2.
XXblocks	List of nxn matrices. Usually output of createXXblocks.
Y	Response vector: numeric, binary, factor or survival.
pairing	Numerical vector of length 3 or NULL when pairs are absent. Represents the indices (in XXblocks) of the two data blocks involved in pairing, plus the index of the paired block.
model	Character. Any of c("linear", "logistic", "cox"). Is inferred from Y when NULL.
reltol	Scalar. Relative tolerance for optimization methods.
optmethod	Character. Optimization method. Any of the methods c("Brent", "Nelder-Mead", "Sann") may be used, but "Nelder-Mead" is generally recommended. Other unconstrained methods offered by optim may also be used, but have not been tested.
maxItropt	Integer. Maximum number of iterations for optmethod.
tracescore	Boolean. Should the output of the scores be traced?
fixedpen	Integer vector or NULL. Contains indices of data types of which penalty is fixed to the corresponding value in penaltiesinit.
fixedseed	Boolean. Should the initialization be fixed? For reproducibility.
sigmasq	Default error variance.
opt.sigma	Boolean. Should the error variance be optimized as well? Only relevant for model="linear".

Details

See gam for details on how the marginal likelihood is computed.

Value

List, with components:

optres	Output of the optimizer
optpen	Vector with determined optimal penalties
allsc	Matrix with marginal likelihood scores for all penalty parameter configurations used by the optimizer

See Also

optLambdas_mgcvWrap for i) (recommended) optimization in two steps: first global, then local; and ii) sequential optimization when some data types are preferred over others. A full demo and data are available from:
https://drive.google.com/open?id=1NUfeOtN8-KZ8A2HZzveG506nBwgW64e4

Examples

data(dataXXmirmeth)
resp <- dataXXmirmeth[[1]]
XXmirmeth <- dataXXmirmeth[[2]]

# Find initial lambdas: fast CV per data block separately.
cvperblock2 <- fastCV2(XXblocks=XXmirmeth,Y=resp,kfold=10,fixedfolds = TRUE)
lambdas <- cvperblock2$lambdas

# Create (repeated) CV-splits of the data.
leftout <- CVfolds(Y=resp,kfold=10,nrepeat=3,fixedfolds = TRUE)

# Compute cross-validated score for initial lambdas
CVscore(penalties=lambdas, XXblocks=XXmirmeth,Y=resp,folds=leftout,
score="loglik")

# Optimize by using marginal likelihood criterion
jointlambdas2 <- optLambdas_mgcvWrap(penaltiesinit=lambdas, XXblocks=XXmirmeth,
Y=resp)

# Optimal lambdas
optlambdas <- jointlambdas2$optpen

multiridge documentation built on June 13, 2022, 5:07 p.m.