Type: Package
Title: Stability-Selection via Correlated Resampling for 'GAMLSS' Models
Version: 0.2.2
Date: 2025-11-20
Depends: R (≥ 4.1.0)
Imports: gamlss, Rcpp, rlang, SelectBoost, stats, utils
Suggests: doParallel, foreach, future, future.apply, gamlss.data, gamlss.dist, ggplot2, glmnet, grpreg, knitr, knockoff, MASS, microbenchmark, nlme, parallel, pkgdown, pscl, rmarkdown, SGL, testthat (≥ 3.0.0)
LinkingTo: Rcpp, RcppArmadillo
Author: Frederic Bertrand ORCID iD [cre, aut]
Maintainer: Frederic Bertrand <frederic.bertrand@lecnam.net>
Description: Extends the 'SelectBoost' approach to Generalized Additive Models for Location, Scale and Shape (GAMLSS). Implements bootstrap stability-selection across parameter-specific formulas (mu, sigma, nu, tau) via gamlss::stepGAIC(). Includes optional standardization of predictors and helper functions for corrected AIC calculation. More details can be found in Bertrand and Maumy (2024) https://hal.science/hal-05352041 that highlights correlation-aware resampling to improve variable selection for GAMLSS and quantile regression when predictors are numerous and highly correlated.
License: GPL-3
Encoding: UTF-8
Classification/MSC: 62H11, 62J12, 62J99
LazyData: true
VignetteBuilder: knitr
RoxygenNote: 7.3.3
URL: https://fbertran.github.io/SelectBoost.gamlss/, https://github.com/fbertran/SelectBoost.gamlss
BugReports: https://github.com/fbertran/SelectBoost.gamlss/issues
Config/testthat/edition: 3
SystemRequirements: C++17
NeedsCompilation: yes
Packaged: 2025-11-21 00:44:35 UTC; bertran7
Repository: CRAN
Date/Publication: 2025-11-25 20:22:11 UTC

SelectBoost.gamlss: Stability-Selection via Correlated Resampling for 'GAMLSS' Models

Description

Extends the 'SelectBoost' approach to Generalized Additive Models for Location, Scale and Shape (GAMLSS). Implements bootstrap stability-selection across parameter-specific formulas (mu, sigma, nu, tau) via gamlss::stepGAIC(). Includes optional standardization of predictors and helper functions for corrected AIC calculation. More details can be found in Bertrand and Maumy (2024) https://hal.science/hal-05352041 that highlights correlation-aware resampling to improve variable selection for GAMLSS and quantile regression when predictors are numerous and highly correlated.

Author(s)

This package was written by Frederic Bertrand. Maintainer: Frederic Bertrand frederic.bertrand@lecnam.net

References

Bertrand, F. and Maumy, M. (2024). An Improvement for Variable Selection for Generalized Additive Models for Location, Shape and Scale and Quantile Regression. Joint Statistics Meetings (JSM) 2024, Portland, OR. Presented the SelectBoost approach for GAMLSS and quantile regression, emphasizing correlation-aware resampling to improve recall and precision when predictors are numerous and highly correlated.

See Also

Useful links:

Examples


set.seed(1)
dat <- data.frame(
  y = gamlss.dist::rNO(80, mu = 0),
  x1 = rnorm(80),
  x2 = rnorm(80)
)
fit <- SelectBoost_gamlss(
  y ~ 1,
  data = dat,
  family = gamlss.dist::NO(),
  mu_scope = ~ x1 + x2,
  B = 10,
  pi_thr = 0.6,
  trace = FALSE
)
fit$final_formula

Reasonable defaults

Description

Adjust as needed per family docs

Usage

.family_defaults()

Value

List of list of default values for parameters for each supported distribution.

Per-family numeric tolerance for equality checks

Description

Per-family numeric tolerance for equality checks

Usage

.family_tolerance()

Value

List of numerical values, the default tolerance, for each supported distribution.

Try to generate values for a family

Description

Try to generate values for a family

Usage

.gen_family(fam, n)

Arguments

fam

Character scalar naming the family. This should correspond to a distribution available in the gamlss.dist package, as used when constructing r<family> generator functions (for example, ⁠NO'' or ⁠LOGNO”).

n

Positive integer giving the number of observations to generate for the requested family.

Value

If successful, a umerical vector of n values randomly generated for the requested family. If the generator is not available or fails, then returns NULL.

AICc for a gamlss fit

Description

AICc for a gamlss fit

Usage

AICc_gamlss(object)

Arguments

object

a 'gamlss' object

Value

numeric AICc value

SelectBoost-style wrapper for GAMLSS

Description

A thin wrapper around sb_gamlss() with SelectBoost-flavored arguments.

Usage

SelectBoost_gamlss(
  formula,
  data,
  family,
  mu_scope,
  sigma_scope = NULL,
  nu_scope = NULL,
  tau_scope = NULL,
  base_sigma = ~1,
  base_nu = ~1,
  base_tau = ~1,
  B = 100,
  sample_fraction = 0.7,
  pi_thr = 0.6,
  k = 2,
  direction = c("both", "forward", "backward"),
  pre_standardize = FALSE,
  use_groups = TRUE,
  c0 = 0.5,
  trace = TRUE,
  ...
)

## S3 method for class 'SelectBoost_gamlss'
summary(object, prop.level = 0.6, ...)

## S3 method for class 'summary.SelectBoost_gamlss'
plot(x, ...)

Arguments

formula

Base formula for the location \mu parameter (in the main model call).

data

Data frame.

family

A gamlss.dist family object (e.g., gamlss.dist::NO()).

mu_scope

Formula of candidate terms for \mu.

sigma_scope, nu_scope, tau_scope

Formulas of candidate terms for \sigma, \nu, \tau.

base_sigma, base_nu, base_tau

Optional base (always-included) formulas for \sigma, \nu, \tau.

B

Number of bootstrap subsamples for stability selection.

sample_fraction

Fraction of rows per subsample (e.g., 0.7).

pi_thr

Selection proportion threshold to define “stable” terms (e.g., 0.6).

k

Penalty weight for stepwise GAIC when engine = "stepGAIC" (default 2).

direction

Stepwise direction for stepGAIC ("both", "forward", "backward").

pre_standardize

Logical; standardize numeric predictors before penalized fits.

use_groups

Logical; enable SelectBoost grouping.

c0

Correlation threshold for grouping (as in SelectBoost::group_func_2).

trace

Logical; print progress messages.

...

Not used.

object

A asb_gamlss object.

prop.level

A target proportion level.

x

A summary of asb_gamlss object.

Value

An object of class c("SelectBoost_gamlss"), with slots similar to sb_gamlss.

A list with selection, threshold and confidence.

Invisibly returns x.

AutoBoost for GAMLSS (SelectBoost-style)

Description

Runs a c0 grid, picks the c0 that maximizes total confidence, and returns the corresponding sb_gamlss fit.

Usage

autoboost_gamlss(
  formula,
  data,
  family,
  mu_scope,
  sigma_scope = NULL,
  nu_scope = NULL,
  tau_scope = NULL,
  base_sigma = ~1,
  base_nu = ~1,
  base_tau = ~1,
  c0_grid = seq(0.1, 0.9, by = 0.1),
  B = 60,
  sample_fraction = 0.7,
  pi_thr = 0.6,
  k = 2,
  direction = c("both", "forward", "backward"),
  pre_standardize = FALSE,
  trace = TRUE,
  progress = TRUE,
  use_groups = TRUE,
  corr_func = "cor",
  group_fun = SelectBoost::group_func_2,
  ...
)

Arguments

formula

Base formula for the location \mu parameter (in the main model call).

data

Data frame.

family

A gamlss.dist family object (e.g., gamlss.dist::NO()).

mu_scope

Formula of candidate terms for \mu.

sigma_scope, nu_scope, tau_scope

Formulas of candidate terms for \sigma, \nu, \tau.

base_sigma, base_nu, base_tau

Optional base (always-included) formulas for \sigma, \nu, \tau.

c0_grid

Numeric vector of c0 values.

B

Number of bootstrap subsamples for stability selection.

sample_fraction

Fraction of rows per subsample (e.g., 0.7).

pi_thr

Selection proportion threshold to define “stable” terms (e.g., 0.6).

k

Penalty weight for stepwise GAIC when engine = "stepGAIC" (default 2).

direction

Stepwise direction for stepGAIC ("both", "forward", "backward").

pre_standardize

Logical; standardize numeric predictors before penalized fits.

trace

Logical; print progress messages.

progress

Logical; show a progress bar in sequential runs.

use_groups

Logical; treat SelectBoost correlation groups during resampling.

corr_func

Correlation function passed to SelectBoost::boost.compcorrs.

group_fun

Grouping function passed to SelectBoost::boost.findgroups.

...

Passed to underlying engines (e.g., to gamlss::gamlss, glmnet, etc.).

Value

A SelectBoost_gamlss_grid with summary plots/tables.

Anthropometric measurements for Dutch boys (1997 reference)

Description

Cross-sectional anthropometric records for 7,482 Dutch boys aged 0 to 21 years that were used to construct the 1997 Dutch growth references. The dataset stores standard auxological indicators together with puberty and regional classification variables.

Format

A data frame with 7,482 rows and 9 variables:

age

Decimal age in years ranging from birth to 21.

hgt

Standing height in centimetres.

wgt

Body weight recorded in kilograms.

bmi

Body mass index.

hc

Head circumference in centimetres.

gen

Ordered factor describing genital Tanner stage (G1–G5).

phb

Ordered factor indicating pubic hair stage (P1–P6).

tv

Testicular volume in millilitres.

reg

Region of residence (north, east, west, south, city).

Details

The table combines the complete cross-sectional sample of Dutch boys aged 0–21 years that formed the basis of the Dutch 1997 growth references. Tanner stage variables are stored as ordered factors, while the region indicator is a nominal factor.

Source

Fredriks, A. M., van Buuren, S., Burgmeijer, R. J., Meulmeester, J. F., Beuker, R. J., Brugman, E., Roede, M. J., Verloove-Vanhorick, S. P., & Wit, J. M. (2000). Continuing positive secular growth change in The Netherlands 1955-1997. Pediatric Research, 47, 316-323.

Fredriks, A. M., van Buuren, S., Wit, J. M., & Verloove-Vanhorick, S. P. (2000). Body index measurements in 1996-97 compared with 1980. Archives of Disease in Childhood, 82, 107-112.

References

Stef van Buuren (2012).

Examples

data(boys7482)
str(boys7482)

Numerical check: fast vs generic deviance log-likelihood

Description

Computes both loglik_gamlss_newdata_fast() and loglik_gamlss_newdata() and reports the absolute difference. Useful for sanity checks.

Usage

check_fast_vs_generic(fit, newdata, tol = 1e-08)

Arguments

fit

A gamlss fit.

newdata

Data frame to evaluate on.

tol

Tolerance for pass/fail (default 1e-8).

Value

A list with fields: ll_fast, ll_generic, abs_diff, pass.

Confidence functionals from a c0 grid

Description

Summarize selection proportions across c0 (SelectBoost threshold) into single-number confidence scores per term/parameter.

Usage

confidence_functionals(
  x,
  pi_thr = NULL,
  q = c(0.5, 0.8, 0.9),
  weight_fun = NULL,
  conservative = FALSE,
  B = NULL,
  method = c("trapezoid", "step")
)

Arguments

x

An object from sb_gamlss_c0_grid().

pi_thr

Stability threshold; defaults to x$pi_thr.

q

Numeric vector of quantiles to compute (in 0..1).

weight_fun

Optional function w(c0) for weighted AUSC; default uniform.

conservative

If TRUE, use Wilson lower confidence bounds for proportions.

B

Number of bootstraps (if not inferable when conservative = TRUE).

method

Integration method: "trapezoid" (default) or "step".

Value

A data.frame with per-term summaries, classed as "sb_confidence".

Compute SelectBoost-like confidence table across c0

Description

Compute SelectBoost-like confidence table across c0

Usage

confidence_table(grid, pi_thr = NULL)

Arguments

grid

an object returned by sb_gamlss_c0_grid

pi_thr

optional override of the threshold (defaults to grid$pi_thr)

Value

data.frame with term, parameter, conf_index (mean positive excess), cover (fraction of c0 with prop>=thr)

K-fold deviance for an sb_gamlss configuration

Description

K-fold deviance for an sb_gamlss configuration

Usage

cv_deviance_sb(K, build_fit, data)

Arguments

K

folds

build_fit

function(...) that returns an sb_gamlss object

data

data.frame used inside build_fit

Value

numeric: mean deviance across folds (-2 * mean loglik)

One-variable effect plot from an sb_gamlss (or gamlss) fit

Description

Varies one variable and holds others at typical values (median/mode) to plot the predicted parameter curve (default: mu). Uses ggplot2 if available, otherwise base.

Usage

effect_plot(fit, var, data, what = "mu", grid = 100)

## S3 method for class 'effect_plot_failure'
print(x, ...)

Arguments

fit

sb_gamlss object (or gamlss)

var

character, name of the variable to vary

data

original data.frame used to fit

what

which parameter to predict ("mu","sigma","nu","tau")

grid

number of grid points for numeric variable

x

object returned by effect_plot() when prediction fails

...

unused

Value

a ggplot object if ggplot2 present; otherwise draws base plot and returns NULL

Invisibly returns x.

Compare fast vs generic deviance log-likelihood evaluation

Description

Compare fast vs generic deviance log-likelihood evaluation

Usage

fast_vs_generic_ll(fit, newdata, reps = 100L, unit = "us")

Arguments

fit

A gamlss fit.

newdata

Data frame to evaluate on.

reps

Number of repetitions (default 100).

unit

microbenchmark unit (default "us").

Value

A data.frame with method, median, and relative speed.

FastBoost for GAMLSS (lightweight stability selection)

Description

A faster variant with fewer bootstraps and smaller subsamples.

Usage

fastboost_gamlss(
  formula,
  data,
  family,
  mu_scope,
  sigma_scope = NULL,
  nu_scope = NULL,
  tau_scope = NULL,
  base_sigma = ~1,
  base_nu = ~1,
  base_tau = ~1,
  B = 30,
  sample_fraction = 0.6,
  pi_thr = 0.6,
  k = 2,
  direction = c("both", "forward", "backward"),
  pre_standardize = FALSE,
  use_groups = TRUE,
  c0 = 0.5,
  trace = TRUE,
  corr_func = "cor",
  group_fun = SelectBoost::group_func_2,
  ...
)

Arguments

formula

Base formula for the location \mu parameter (in the main model call).

data

Data frame.

family

A gamlss.dist family object (e.g., gamlss.dist::NO()).

mu_scope

Formula of candidate terms for \mu.

sigma_scope, nu_scope, tau_scope

Formulas of candidate terms for \sigma, \nu, \tau.

base_sigma, base_nu, base_tau

Optional base (always-included) formulas for \sigma, \nu, \tau.

B

Number of bootstrap subsamples for stability selection.

sample_fraction

Fraction of rows per subsample (e.g., 0.7).

pi_thr

Selection proportion threshold to define “stable” terms (e.g., 0.6).

k

Penalty weight for stepwise GAIC when engine = "stepGAIC" (default 2).

direction

Stepwise direction for stepGAIC ("both", "forward", "backward").

pre_standardize

Logical; standardize numeric predictors before penalized fits.

use_groups

Logical; treat SelectBoost correlation groups during resampling.

c0

SelectBoost meta-parameter controlling reweighting/thresholding (see vignette).

trace

Logical; print progress messages.

corr_func

Correlation function passed to SelectBoost::boost.compcorrs.

group_fun

Grouping function passed to SelectBoost::boost.findgroups.

...

Passed to underlying engines (e.g., to gamlss::gamlss, glmnet, etc.).

Details

Fast SelectBoost (single c0)

Value

An sb_gamlss fit at the given c0.

Get a density function for a gamlss family

Description

Get a density function for a gamlss family

Usage

get_density_fun(fit)

Arguments

fit

a gamlss fit (or family name)

Value

function(x, mu, sigma, nu, tau, log=FALSE)

Knockoff filter for mu (approximate group control)

Description

Knockoff filter for mu (approximate group control)

Usage

knockoff_filter_mu(data, response, mu_scope, fdr = 0.1, df_smooth = 6L)

Arguments

data

data.frame

response

response variable name

mu_scope

RHS-only term labels

fdr

target FDR level

df_smooth

df for smoother proxies (splines::bs)

Value

character vector of selected term names

Knockoff filter for sigma/nu/tau (approximate group control)

Description

Knockoff filter for sigma/nu/tau (approximate group control)

Usage

knockoff_filter_param(data, scope, y_work, fdr = 0.1, df_smooth = 6L)

Arguments

data

data.frame

scope

RHS-only term labels

y_work

working response (numeric)

fdr

target FDR level

df_smooth

df for smoother proxies

Value

character vector of selected term names

Log-likelihood (sum) on newdata given a gamlss fit

Description

Log-likelihood (sum) on newdata given a gamlss fit

Usage

loglik_gamlss_newdata(fit, newdata)

Arguments

fit

gamlss object

newdata

data.frame

Value

numeric scalar: sum of log-likelihoods

Plot selection proportions for a single sb_gamlss

Description

Plot selection proportions for a single sb_gamlss

Usage

## S3 method for class 'SelectBoost_gamlss'
plot(x, ...)

Arguments

x

A sb_gamlss object.

...

Graphical parameters.

Value

Invisibly returns x the plotted sb_gamlss object.

Plot summary for sb_gamlss_c0_grid

Description

Plot summary for sb_gamlss_c0_grid

Usage

## S3 method for class 'SelectBoost_gamlss_grid'
plot(x, top = 15, ...)

Arguments

x

A SelectBoost_gamlss_grid object.

top

Integer; how many top terms to show in the confidence barplot.

...

Ignored (reserved for future).

Value

An invisible copy of x.

Plot confidence functionals

Description

Two-panel plot: (1) scatter of area_pos vs cover (size by rank), (2) barplot of top-N rank_score.

Usage

## S3 method for class 'sb_confidence'
plot(x, top = 15, label_top = 10, ...)

Arguments

x

An object from confidence_functionals().

top

Show top-N terms in the barplot (default 15).

label_top

Integer; number of points to label in the scatter (default 10).

...

Graphical parameters passed to plotting backend.

Value

An invisible copy of x.

Plot selection frequencies for sb_gamlss

Description

Plot selection frequencies for sb_gamlss

Usage

plot_sb_gamlss(x, top = Inf, ...)

Arguments

x

A sb_gamlss object

top

Show only the top N terms per-parameter (default all)

...

Graphical parameters.

Value

Invisibly returns x the plotted sb_gamlss object.

Plot stability curves p(c0) for selected terms

Description

Plot stability curves p(c0) for selected terms

Usage

plot_stability_curves(grid, terms, parameter = NULL, ncol = 2L)

Arguments

grid

An object from sb_gamlss_c0_grid().

terms

Character vector of term names to plot.

parameter

Optional parameter name ('mu','sigma','nu','tau'); if NULL, all.

ncol

Columns in the multi-panel layout.

Value

Invisibly returns grid the plotted object.

Predict distribution parameters on newdata

Description

Predict distribution parameters on newdata

Usage

predict_params(fit, newdata)

Arguments

fit

a gamlss fit

newdata

data.frame

Value

list with available components: mu, sigma, nu, tau

SelectBoost for GAMLSS (stability selection)

Description

SelectBoost for GAMLSS (stability selection)

Usage

sb_gamlss(
  formula,
  data,
  family,
  mu_scope,
  sigma_scope = NULL,
  nu_scope = NULL,
  tau_scope = NULL,
  base_sigma = ~1,
  base_nu = ~1,
  base_tau = ~1,
  B = 100,
  sample_fraction = 0.7,
  pi_thr = 0.6,
  k = 2,
  direction = c("both", "forward", "backward"),
  pre_standardize = FALSE,
  use_groups = FALSE,
  c0 = 0.5,
  engine = c("stepGAIC", "glmnet", "grpreg", "sgl"),
  engine_sigma = NULL,
  engine_nu = NULL,
  engine_tau = NULL,
  grpreg_penalty = c("grLasso", "grMCP", "grSCAD"),
  sgl_alpha = 0.95,
  df_smooth = 6L,
  progress = TRUE,
  glmnet_alpha = 1,
  glmnet_family = c("gaussian", "binomial", "poisson"),
  parallel = c("none", "auto", "multisession", "multicore"),
  workers = NULL,
  trace = TRUE,
  corr_func = "cor",
  group_fun = SelectBoost::group_func_2,
  ...
)

Arguments

formula

Base formula for the location \mu parameter (in the main model call).

data

Data frame.

family

A gamlss.dist family object (e.g., gamlss.dist::NO()).

mu_scope

Formula of candidate terms for \mu.

sigma_scope, nu_scope, tau_scope

Formulas of candidate terms for \sigma, \nu, \tau.

base_sigma, base_nu, base_tau

Optional base (always-included) formulas for \sigma, \nu, \tau.

B

Number of bootstrap subsamples for stability selection.

sample_fraction

Fraction of rows per subsample (e.g., 0.7).

pi_thr

Selection proportion threshold to define “stable” terms (e.g., 0.6).

k

Penalty weight for stepwise GAIC when engine = "stepGAIC" (default 2).

direction

Stepwise direction for stepGAIC ("both", "forward", "backward").

pre_standardize

Logical; standardize numeric predictors before penalized fits.

use_groups

Logical; treat SelectBoost correlation groups during resampling.

c0

SelectBoost meta-parameter controlling reweighting/thresholding (see vignette).

engine

Engine for \mu ("stepGAIC", "glmnet", "grpreg", "sgl").

engine_sigma, engine_nu, engine_tau

Optional engines for \sigma, \nu, \tau.

grpreg_penalty

Group penalty for grpreg ("grLasso", "grMCP", "grSCAD").

sgl_alpha

Alpha for sparse group lasso.

df_smooth

Degrees of freedom for proxy spline bases (pb()/cs() → splines::bs(df=df_smooth)) used only for grouped selection design.

progress

Logical; show a progress bar in sequential runs.

glmnet_alpha

Elastic-net mixing for glmnet (1 = lasso, 0 = ridge).

glmnet_family

Family passed to glmnet-based selectors ("gaussian", "binomial", "poisson").

parallel

Parallel mode ("none", "auto", "multisession", '"multicore").

workers

Integer; number of workers if parallel.

trace

Logical; print progress messages.

corr_func

Correlation function passed to SelectBoost::boost.compcorrs.

group_fun

Grouping function passed to SelectBoost::boost.findgroups.

...

Passed to underlying engines (e.g., to gamlss::gamlss, glmnet, etc.).

Value

An object of class "sb_gamlss" with elements:

Examples


set.seed(1)
dat <- data.frame(
  y = gamlss.dist::rNO(60, mu = 0),
  x1 = rnorm(60),
  x2 = rnorm(60),
  x3 = rnorm(60)
)
fit <- sb_gamlss(
  y ~ 1,
  data = dat,
  family = gamlss.dist::NO(),
  mu_scope = ~ x1 + x2 + gamlss::pb(x3),
  B = 8,
  pi_thr = 0.6,
  trace = FALSE
)
fit$final_formula

Stability curves over a c0 grid for sb_gamlss

Description

Stability curves over a c0 grid for sb_gamlss

Usage

sb_gamlss_c0_grid(
  formula,
  data,
  family,
  mu_scope,
  sigma_scope = NULL,
  nu_scope = NULL,
  tau_scope = NULL,
  base_sigma = ~1,
  base_nu = ~1,
  base_tau = ~1,
  c0_grid = seq(0.1, 0.9, by = 0.1),
  B = 60,
  sample_fraction = 0.7,
  pi_thr = 0.6,
  k = 2,
  direction = c("both", "forward", "backward"),
  pre_standardize = FALSE,
  trace = TRUE,
  progress = TRUE,
  use_groups = TRUE,
  corr_func = "cor",
  group_fun = SelectBoost::group_func_2,
  ...
)

Arguments

formula

Base formula for the location \mu parameter (in the main model call).

data

Data frame.

family

A gamlss.dist family object (e.g., gamlss.dist::NO()).

mu_scope

Formula of candidate terms for \mu.

sigma_scope, nu_scope, tau_scope

Formulas of candidate terms for \sigma, \nu, \tau.

base_sigma, base_nu, base_tau

Optional base (always-included) formulas for \sigma, \nu, \tau.

c0_grid

Numeric vector of c0 thresholds in (0,1).

B

Number of bootstrap subsamples for stability selection.

sample_fraction

Fraction of rows per subsample (e.g., 0.7).

pi_thr

Selection proportion threshold to define “stable” terms (e.g., 0.6).

k

Penalty weight for stepwise GAIC when engine = "stepGAIC" (default 2).

direction

Stepwise direction for stepGAIC ("both", "forward", "backward").

pre_standardize

Logical; standardize numeric predictors before penalized fits.

trace

Logical; print progress messages.

progress

Logical; show a progress bar across c0_grid.

use_groups

Logical; treat SelectBoost correlation groups during resampling.

corr_func

Correlation function passed to SelectBoost::boost.compcorrs.

group_fun

Grouping function passed to SelectBoost::boost.findgroups.

...

Passed to underlying engines (e.g., to gamlss::gamlss, glmnet, etc.).

Value

An object of class "SelectBoost_gamlss_grid" containing

Selection table accessor

Description

Selection table accessor

Usage

selection_table(x)

Arguments

x

A sb_gamlss object

Value

data.frame with parameter, term, count, prop

Tune select engines/penalties via a small stability run

Description

Evaluates a grid of configurations and picks the one maximizing a stability-based score, optionally penalized by complexity. Designed to be lightweight and robust.

Usage

tune_sb_gamlss(
  config_grid,
  base_args,
  score_lambda = 0,
  B_small = 30,
  metric = c("stability", "deviance"),
  K = 3,
  progress = TRUE
)

Arguments

config_grid

a list of named lists, each containing a subset of sb_gamlss args (e.g., list(engine="grpreg", engine_sigma="sgl", grpreg_penalty="grLasso", sgl_alpha=0.9))

base_args

a named list of arguments passed to sb_gamlss() common to all configs

score_lambda

Numeric; complexity penalty weight for stability metric.

B_small

number of bootstraps to use during tuning (defaults to 30)

metric

Character; "stability" or "deviance" (K-fold CV).

K

Integer; folds for deviance CV.

progress

Logical; show progress bar across configs.

Value

a list: best_config, scores (data.frame), and the fitted sb_gamlss object for the best config.