Package 'S3VS' reference manual

Title:	Structured Screen-and-Select Variable Selection in Linear, Generalized Linear, and Survival Models
Description:	Performs variable selection using the structured screen-and-select (S3VS) framework in linear models, generalized linear models with binary data, and survival models such as the Cox model and accelerated failure time (AFT) model.
Authors:	Nilotpal Sanyal [aut, cre], Padmore N. Prempeh [aut]
Maintainer:	Nilotpal Sanyal <[email protected]>
License:	GPL (>= 2)
Version:	1.0
Built:	2026-05-29 19:26:33 UTC
Source:	https://github.com/cran/S3VS

Structured Screen-and-Select Variable Selection in Linear, Generalized Linear, and Survival Models

Description

Performs variable selection using the structured screen-and-select (S3VS) framework in linear models, generalized linear models with binary data, and survival models such as the Cox model and accelerated failure time (AFT) model.

Details

The S3VS package implements the Structured Screen-and-Select Variable Selection (S3VS) framework for linear models, generalized linear models with binary responses, and survival models (Cox proportional hazards and accelerated failure time models).

The central entry point is S3VS, which dispatches to a family-specific routine via the argument family:

S3VS_LM for linear models,
S3VS_GLM for generalized linear models with binary outcomes,
S3VS_SURV for survival models.

The S3VS workflow proceeds through the following steps, each handled by helper functions:

Stopping rule check: looprun determines whether the iterative screen-and-select process should continue.
Leading variable identification: get_leadvars identifies leading variables; family-specific versions are get_leadvars_LM, get_leadvars_GLM, and get_leadvars_SURV.
Leading set identification: get_leadsets identifies the leading set for each leading variable.
Selection within leading sets: VS_method performs selection within leading sets; family-specific methods include VS_method_LM, VS_method_GLM, VS_method_SURV, and bridge_aft implements BRIDGE specifically for AFT models.
Aggregation of selected variables: select_vars retains promising variables as selected from an iteration.
Aggregation of non-selected variables (optional): remove_vars removes variables deemed uninformative from future iterations (if no variable is selected in the current iteration by select_vars).
Response update (optional): update_y enables iterative response updates; family-specific variants include update_y_LM and update_y_GLM.

Together, these functions form a structured, iterative pipeline for efficient variable screening and selection in high-dimensional regression and survival analysis.

Prediction: pred_S3VS produces predictions using variables selected by S3VS, calling pred_S3VS_LM, pred_S3VS_GLM, or pred_S3VS_SURV as appropriate.

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Maintainer: Nilotpal Sanyal <[email protected]>

Bridge-Penalized AFT Regression via Iteratively Reweighted LASSO

Description

bridge_aft fits an accelerated failure time (AFT) model using an iterative reweighted LASSO scheme to approximate a bridge ( $L_\gamma$ ) penalty on the regression coefficients.

Usage

bridge_aft(y, X, gamma = 0.5, alpha = 1, max_iter = 100, tol = 1e-05)
bridge_aft(y, X, gamma = 0.5, alpha = 1, max_iter = 100, tol = 1e-05)

Arguments

y

Response; a list of two elements time and status.

X

Predictor matrix. Can be a base matrix or something as.matrix() can coerce. No missing values are allowed.

gamma

Bridge penalty exponent $\gamma$ (default 0.5). Values in $(0,1]$ approximate nonconvex penalties; $\gamma=1$ reduces to LASSO-like weighting.

alpha

Elastic-net mixing parameter passed to glmnet::cv.glmnet (default 1 for LASSO; 0 is ridge; (0,1) is elastic net).

max_iter

Maximum number of outer reweighting iterations (default 100).

tol

Convergence tolerance on the $L_1$ change in coefficients between iterations (default 1e-5).

Value

A list with components:

beta

Numeric vector of estimated coefficients of length $p$ .

gamma

The bridge exponent used in the fit.

iterations

Number of outer reweighting iterations performed.

Author(s)

Padmore Prempeh <[email protected]>, Nilotpal Sanyal <[email protected]>

References

Jian Huang and Shuangge Ma. Variable selection in the accelerated failure time model via the bridge method. Lifetime Data Analysis, 16(2):176-195, 2010.

Examples

set.seed(1)
n  <- 50
p  <- 10
X  <- matrix(rnorm(n * p), n, p)
beta_true <- c(runif(10, -1.5, 1.5), rep(0, p - 10))
linpred   <- as.vector(X %*% beta_true)

## Generate log-normal AFT survival times (no censoring in this simple example)
sigma <- 0.6
logT  <- linpred + rnorm(n, sd = sigma)
time  <- exp(logT)
delta <- rep(1, n)  # all events (censoring ignored by current implementation)

y_surv <- list(time = time, status = delta)
fit <- bridge_aft(y_surv, X, gamma = 0.5, alpha = 1, max_iter = 50, tol = 1e-5)
str(fit)
fit$beta[1:10]
set.seed(1)
n  <- 50
p  <- 10
X  <- matrix(rnorm(n * p), n, p)
beta_true <- c(runif(10, -1.5, 1.5), rep(0, p - 10))
linpred   <- as.vector(X %*% beta_true)

## Generate log-normal AFT survival times (no censoring in this simple example)
sigma <- 0.6
logT  <- linpred + rnorm(n, sd = sigma)
time  <- exp(logT)
delta <- rep(1, n)  # all events (censoring ignored by current implementation)

y_surv <- list(time = time, status = delta)
fit <- bridge_aft(y_surv, X, gamma = 0.5, alpha = 1, max_iter = 50, tol = 1e-5)
str(fit)
fit$beta[1:10]

Identify Leading Sets of Covariates via Inter-Predictor Associations

Description

get_leadsets identifies, for a specified leading variable, a set of associated predictors, the leading set, based on inter-predictor associations (absolute value of the correlation coefficient).

Usage

get_leadsets(x_lead, X, method = c("topk", "fixedthresh", "percthresh"), param)
get_leadsets(x_lead, X, method = c("topk", "fixedthresh", "percthresh"), param)

Arguments

x_lead

Vector with values of the leading variable

X

Predictor matrix. Must contain the leading variable. Can be a base matrix or something as.matrix() can coerce. No missing values are allowed.

method

Rule for constructing, for each leading variable, the set of associated predictors (the "leading set") using inter-predictor association (absolute value of the correlation coefficient); one of c("topk", "fixedthresh", "percthresh"). "topk" keeps the predictors with the largest $k$ association values; "fixedthresh" keeps predictors whose association is greater than or equal to a specified threshold; "percthresh" keeps predictors whose association is within a given percentage of the best.

param

Tuning parameter for method; If "topk", supply an integer $k$ (keep the top $k$ ). If "fixedthresh", supply a numeric threshold (keep predictors with association $\ge$ threshold). If "percthresh", supply a percentage in $(0,100]$ (keep predictors with association $\ge$ that percent of the highest association).

Value

A character vector containing the names of the predictors.

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

# Simulate continuous data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
y <- X[,1] + 0.5 * X[,2] + rnorm(n)
leadvars <- get_leadvars_LM(y = y, X = X, method = "topk", param = list(k=2))
get_leadsets(X[,leadvars[1]], X, method = "percthresh", param = list(thresh = 0.2))
# Simulate continuous data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
y <- X[,1] + 0.5 * X[,2] + rnorm(n)
leadvars <- get_leadvars_LM(y = y, X = X, method = "topk", param = list(k=2))
get_leadsets(X[,leadvars[1]], X, method = "percthresh", param = list(thresh = 0.2))

Screening Predictors As 'Leading Variables' By Evaluating Predictor-Response Associations

Description

get_leadvars screens some predictors as "leading variables" based on predictor-response associations in linear, generalized linear, and survival models.

Usage

get_leadvars(y, X, family = c("normal","binomial","survival"), 
  surv_model = c("AFT", "COX"), 
  method = c("topk", "fixedthresh", "percthresh"), param, 
  varsselected = NULL, varsleft = colnames(X), parallel = FALSE)
get_leadvars(y, X, family = c("normal","binomial","survival"), 
  surv_model = c("AFT", "COX"), 
  method = c("topk", "fixedthresh", "percthresh"), param, 
  varsselected = NULL, varsleft = colnames(X), parallel = FALSE)

Arguments

y

Response. If family = "normal", a numeric vector. If family = "binomial", a numeric/integer/logical vector with values in {0,1}. If family = "survival", a list with components time and status (1 = event, 0 = censored).

X

Predictor matrix. Can be a base matrix or something as.matrix() can coerce. No missing values are allowed.

family

Model family; one of c("normal","binomial","survival"). Determines which engine is called (get_leadvars_LM, get_leadvars_GLM, or get_leadvars_SURV).

surv_model

Character string specifying the survival model (family="survival" only). Must be explicitly provided; there is no default. Values are "Cox" for proportional hazards models, "AFT" for accelerated failure time models.

method

Screening rule, one of c("topk", "fixedthresh", "percthresh"). The association measure depends on family (e.g., correlation for "normal", eta-squared for "binomial", or marginal utility for "survival"). "topk" keeps the predictors with the largest $k$ association values; "fixedthresh" keeps predictors whose association is greater than or equal to a specified threshold; "percthresh" keeps predictors whose association is within a given percentage of the best.

param

Tuning parameter for method. If "topk", supply an integer $k$ (keep the top $k$ ). If "fixedthresh", supply a numeric threshold (keep predictors with association $\ge$ threshold). If "percthresh", supply a percentage in $(0,100]$ (keep predictors with association $\ge$ that percent of the highest association).

varsselected

Used only when family=survival. A character vector containing the predictors that are already selected in previous iterations. The association measure, conditional utility, is computed controling for these predictors. NULL, by default.

varsleft

Used only when family=survival. A character vector containing the predictors that are neither selected, nor removed from consideration in previous iterations. Leading predictors are chosen from these predictors. colnames(X), by default.

parallel

Logical. If TRUE, attempts to perform some computations in parallel mode in binomial and survival families, which is strongly recommended for faster execution. Defaults to FALSE.

Value

A character vector containing the names of the leading variables.

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

# Simulate continuous data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
y <- X[,1] + 0.5 * X[,2] + rnorm(n)
# Select leading variables
leadvars <- get_leadvars(y = y, X = X, family = "normal", 
                         method = "topk", param = list(k=2))
leadvars
# Simulate continuous data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
y <- X[,1] + 0.5 * X[,2] + rnorm(n)
# Select leading variables
leadvars <- get_leadvars(y = y, X = X, family = "normal", 
                         method = "topk", param = list(k=2))
leadvars

Screening Predictors As 'Leading Variables' By Evaluating Predictor-Response Associations In Generalized Linear Models

Description

get_leadvars_GLM screens some predictors as "leading variables" based on predictor-response associations in generalized linear models.

Usage

get_leadvars_GLM(y, X, method = c("topk", "fixedetasqthresh", "percetasqthresh"), param)
get_leadvars_GLM(y, X, method = c("topk", "fixedetasqthresh", "percetasqthresh"), param)

Arguments

y

Response. A numeric/integer/logical vector with values in {0,1}.

X

Predictor matrix. Can be a base matrix or something as.matrix() can coerce. No missing values are allowed.

method

Screening rule, one of c("topk", "fixedthresh", "percthresh"). The association measure is eta-squared. "topk" keeps the predictors with the largest $k$ association values; "fixedthresh" keeps predictors whose association is greater than or equal to a specified threshold; "percthresh" keeps predictors whose association is within a given percentage of the best.

param

Value

A character vector containing the names of the leading varibales.

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

# Simulate binary data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
eta <- X[,1] + 0.5 * X[,2]
prob <- 1 / (1 + exp(-eta))
y <- rbinom(n, size = 1, prob = prob)
# Select leading variables
leadvars <- get_leadvars_GLM(y = y, X = X, method = "topk", param = list(k=2))
leadvars
# Simulate binary data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
eta <- X[,1] + 0.5 * X[,2]
prob <- 1 / (1 + exp(-eta))
y <- rbinom(n, size = 1, prob = prob)
# Select leading variables
leadvars <- get_leadvars_GLM(y = y, X = X, method = "topk", param = list(k=2))
leadvars

Screening Predictors As 'Leading Variables' By Evaluating Predictor-Response Associations In Linear Models

Description

get_leadvars_LM screens some predictors as "leading variables" based on predictor-response associations in linear models.

Usage

get_leadvars_LM(y, X, method = c("topk", "fixedcorthresh", "perccorthresh"), param)
get_leadvars_LM(y, X, method = c("topk", "fixedcorthresh", "perccorthresh"), param)

Arguments

y

Response. A numeric vector.

X

Predictor matrix. Can be a base matrix or something as.matrix() can coerce. No missing values are allowed.

method

Screening rule, one of c("topk", "fixedthresh", "percthresh"). The association measure is correlation. "topk" keeps the predictors with the largest $k$ association values; "fixedthresh" keeps predictors whose association is greater than or equal to a specified threshold; "percthresh" keeps predictors whose association is within a given percentage of the best.

param

Value

A character vector containing the names of the leading varibales.

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

# Simulate continuous data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
y <- X[,1] + 0.5 * X[,2] + rnorm(n)
# Select leading variables
leadvars <- get_leadvars_LM(y = y, X = X, method = "topk", param = list(k=2))
leadvars
# Simulate continuous data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
y <- X[,1] + 0.5 * X[,2] + rnorm(n)
# Select leading variables
leadvars <- get_leadvars_LM(y = y, X = X, method = "topk", param = list(k=2))
leadvars

Screening Predictors As "Leading Variables" By Evaluating Predictor-Response Associations In Survival Models

Description

get_leadvars_SURV screens some predictors as "leading variables" based on predictor-response associations in survival models.

Usage

get_leadvars_SURV(y, X, surv_model = c("AFT", "COX"), 
  method = c("topk", "fixedmuthresh", "percmuthresh"), param, 
  varsselected = NULL, varsleft = colnames(X), parallel = FALSE)
get_leadvars_SURV(y, X, surv_model = c("AFT", "COX"), 
  method = c("topk", "fixedmuthresh", "percmuthresh"), param, 
  varsselected = NULL, varsleft = colnames(X), parallel = FALSE)

Arguments

y

Response. A list with components time and status (1 = event, 0 = censored).

X

Predictor matrix. Can be a base matrix or something as.matrix() can coerce. No missing values are allowed.

surv_model

Character string specifying the survival model. Must be explicitly provided; there is no default. Values are "Cox" for proportional hazards models, "AFT" for accelerated failure time models.

method

Screening rule, one of c("topk", "fixedthresh", "percthresh"). The association measure is marginal utility. "topk" keeps the predictors with the largest $k$ association values; "fixedthresh" keeps predictors whose association is greater than or equal to a specified threshold; "percthresh" keeps predictors whose association is within a given percentage of the best.

param

varsselected

A character vector containing the predictors that are already selected in previous iterations. The association measure, conditional utility, is computed controling for these predictors. NULL, by default.

varsleft

A character vector containing the predictors that are neither selected, nor removed from consideration in previous iterations. Leading predictors are chosen from these predictors.

parallel

Logical. If TRUE, attempts to perform some computations in parallel mode in binomial and survival families, which is strongly recommended for faster execution. Defaults to colnames(X).

Value

A character vector containing the names of the leading varibales.

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

# Simulate survival data (Cox)
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
eta <- X[,1] + 0.5 * X[,2]
base_rate <- 0.05
T_event <- rexp(n, rate = base_rate * exp(eta))
C <- rexp(n, rate = 0.03)
time <- pmin(T_event, C)
status <- as.integer(T_event <= C)
y_surv <- list(time = time, status = status)
# Select leading variables
leadvars <- get_leadvars_SURV(y = y_surv, X = X, surv_model = "COX", 
                              method = "topk", param = list(k=2), 
                              varsselected = NULL, varsleft = colnames(X))
leadvars
# Simulate survival data (Cox)
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
eta <- X[,1] + 0.5 * X[,2]
base_rate <- 0.05
T_event <- rexp(n, rate = base_rate * exp(eta))
C <- rexp(n, rate = 0.03)
time <- pmin(T_event, C)
status <- as.integer(T_event <= C)
y_surv <- list(time = time, status = status)
# Select leading variables
leadvars <- get_leadvars_SURV(y = y_surv, X = X, surv_model = "COX", 
                              method = "topk", param = list(k=2), 
                              varsselected = NULL, varsleft = colnames(X))
leadvars

Decide Whether to Run Another S3VS Iteration

Description

looprun evaluates simple stopping criteria for the S3VS procedure and returns an indicator of whether one more iteration should be executed.

Usage

looprun(varsselected, varsleft, max_nocollect, m, nskip)
looprun(varsselected, varsleft, max_nocollect, m, nskip)

Arguments

varsselected

Character vector with names of predictors selected so far. Only its length is used; NULL is treated as length 0.

varsleft

Character vector with names of candidate predictors that remain available for selection in future iterations. Only its length is used; NULL is treated as length 0.

max_nocollect

Integer count of iterations up to now in which no new predictors were selected.

m

Maximum allowed number of selected predictors (target cap for length(varsselected)).

nskip

Maximum allowed number of "no-collection" iterations before stopping.

Details

An additional S3VS iteration is recommended iff all three conditions hold:

$|\texttt{varsselected}| < m,$

$|\texttt{varsleft}| > 0,$

$\texttt{max\_nocollect} < \texttt{nskip}.$

Value

1 if another iteration should run, 0 otherwise.

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

looprun(varsselected = c("x1","x2","x3"),
        varsleft     = paste0("x", 4:23),
        max_nocollect = 0,
        m = 10,
        nskip = 2)
looprun(varsselected = c("x1","x2","x3"),
        varsleft     = paste0("x", 4:23),
        max_nocollect = 0,
        m = 10,
        nskip = 2)

Prediction Using S3VS-Selected Predictors

Description

pred_S3VS performs prediction using predictors selected by S3VS in linear, generalized linear, and survival models.

Usage

pred_S3VS(y, X, family, surv_model = NULL, method)
pred_S3VS(y, X, family, surv_model = NULL, method)

Arguments

y

X

Predictor matrix. This should include predictors selected by S3VS. Can be a base matrix or something as.matrix() can coerce. No missing values are allowed.

family

Model family; one of c("normal","binomial","survival"). Determines which engine is called (pred_LM, pred_GLM, or pred_SURV).

surv_model

method

Character string indicating the prediction method used. Allowed values depend on family: for family = "normal"(functions pred_S3VS_LM), available options are "NLP", "LASSO", "SCAD", "MCP"; for "binomial" (S3VS_GLM), available options are "NLP", "LASSO"; for family = "survival" (S3VS_SURV), available options are "COXGLMNET" for surv_model = "COX" and .... for surv_model = "AFT". See Details for more information.

Value

A list containing:

y.pred

Predicted response

coef

Coefficient estimates of the predictors used for prediction

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

# Simulate continuous data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
y <- X[,1] + 0.5 * X[,2] + rnorm(n)
# Run S3VS for LM
res_lm <- S3VS(y = y, X = X, family = "normal",
               method_xy = "topk", param_xy = list(k=1),
               method_xx = "topk", param_xx = list(k=3),
               vsel_method = "LASSO", method_sel = "conservative", 
               method_rem = "conservative_begin", rem_regout = FALSE, 
               m = 100, nskip = 3, verbose = TRUE, seed = 123)
pred_lm <- pred_S3VS(y = y, X = X[,res_lm$selected], family = "normal", method = "LASSO")
# Simulate continuous data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
y <- X[,1] + 0.5 * X[,2] + rnorm(n)
# Run S3VS for LM
res_lm <- S3VS(y = y, X = X, family = "normal",
               method_xy = "topk", param_xy = list(k=1),
               method_xx = "topk", param_xx = list(k=3),
               vsel_method = "LASSO", method_sel = "conservative", 
               method_rem = "conservative_begin", rem_regout = FALSE, 
               m = 100, nskip = 3, verbose = TRUE, seed = 123)
pred_lm <- pred_S3VS(y = y, X = X[,res_lm$selected], family = "normal", method = "LASSO")

Prediction Using S3VS-Selected Predictors in Survival Models

Description

pred_S3VS performs prediction using predictors selected by S3VS in survival models.

Usage

pred_S3VS_GLM(y, X, method = c("NLP", "LASSO"))
pred_S3VS_GLM(y, X, method = c("NLP", "LASSO"))

Arguments

y

Response. A numeric/integer/logical vector with values in {0,1}.

X

Predictor matrix. This should include predictors selected by S3VS. Can be a base matrix or something as.matrix() can coerce. No missing values are allowed.

method

Character string indicating the prediction method used. Available options are "NLP", "LASSO".

Value

A list containing:

y.pred

Predicted response

coef

Coefficient estimates of the predictors used for prediction

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

# Simulate binary data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
eta <- X[,1] + 0.5 * X[,2]
prob <- 1 / (1 + exp(-eta))
y <- rbinom(n, size = 1, prob = prob)
# Predict
pred_glm <- pred_S3VS_GLM(y = y, X = X[,1:3], method = "LASSO")
pred_glm
# Simulate binary data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
eta <- X[,1] + 0.5 * X[,2]
prob <- 1 / (1 + exp(-eta))
y <- rbinom(n, size = 1, prob = prob)
# Predict
pred_glm <- pred_S3VS_GLM(y = y, X = X[,1:3], method = "LASSO")
pred_glm

Prediction Using S3VS-Selected Predictors in Linear Models

Description

pred_S3VS performs prediction using predictors selected by S3VS in linear models.

Usage

pred_S3VS_LM(y, X, method)
pred_S3VS_LM(y, X, method)

Arguments

y

Response. A numeric vector.

X

Predictor matrix. This should include predictors selected by S3VS. Can be a base matrix or something as.matrix() can coerce. No missing values are allowed.

method

Character string indicating the prediction method used. Available options are "NLP", "LASSO", "SCAD", "MCP"

Value

A list containing:

y.pred

Predicted response

coef

Coefficient estimates of the predictors used for prediction

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

# Simulate continuous data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
y <- X[,1] + 0.5 * X[,2] + rnorm(n)
# Run S3VS for LM
res_lm <- S3VS(y = y, X = X, family = "normal",
               method_xy = "topk", param_xy = list(k=1),
               method_xx = "topk", param_xx = list(k=3),
               vsel_method = "LASSO", method_sel = "conservative", 
               method_rem = "conservative_begin", rem_regout = FALSE, 
               m = 100, nskip = 3, verbose = TRUE, seed = 123)
pred_lm <- pred_S3VS_LM(y = y, X = X[,res_lm$selected], method = "LASSO")
pred_lm
# Simulate continuous data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
y <- X[,1] + 0.5 * X[,2] + rnorm(n)
# Run S3VS for LM
res_lm <- S3VS(y = y, X = X, family = "normal",
               method_xy = "topk", param_xy = list(k=1),
               method_xx = "topk", param_xx = list(k=3),
               vsel_method = "LASSO", method_sel = "conservative", 
               method_rem = "conservative_begin", rem_regout = FALSE, 
               m = 100, nskip = 3, verbose = TRUE, seed = 123)
pred_lm <- pred_S3VS_LM(y = y, X = X[,res_lm$selected], method = "LASSO")
pred_lm

Predicted Survival Probabilities Using S3VS-Selected Predictors in Generalized Linear Models

Description

pred_S3VS returns predicted survival probabilities using predictors selected by S3VS in generalized linear models.

Usage

pred_S3VS_SURV(y, X, surv_model = c("AFT", "COX"), method = c("AFTREG", "AFTGEE"), times)
pred_S3VS_SURV(y, X, surv_model = c("AFT", "COX"), method = c("AFTREG", "AFTGEE"), times)

Arguments

y

Response. A list with components time and status (1 = event, 0 = censored).

X

Predictor matrix. This should include predictors selected by S3VS. Can be a base matrix or something as.matrix() can coerce. No missing values are allowed.

surv_model

Character string specifying the survival model. Must be explicitly provided; there is no default. Values are "Cox" for proportional hazards models, "AFT" for accelerated failure time models.

method

Character string indicating the prediction method used. Available options are "COXGLMNET" for surv_model = "COX" and "AFTREG" and "AFTGEE" for surv_model = "AFT". See Details for more information.

times

Vector of time points where predicted survival probabilities will be computed.

Value

A list containing:

y.pred

Predicted response

coef

Coefficient estimates of the predictors used for prediction

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

# Simulate survival data (Cox)
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
eta <- X[,1] + 0.5 * X[,2]
base_rate <- 0.05
T_event <- rexp(n, rate = base_rate * exp(eta))
C <- rexp(n, rate = 0.03)
time <- pmin(T_event, C)
status <- as.integer(T_event <= C)
y_surv <- list(time = time, status = status)
# Run S3VS for linear models
res_surv <- S3VS(y = y_surv, X = X, family = "survival", 
                 surv_model = "COX", vsel_method = "COXGLMNET",
                 method_xy = "topk", param_xy = list(k = 1),
                 method_xx = "topk", param_xx = list(k = 3),
                 method_sel = "conservative", method_rem = "conservative_begin",
                 sel_regout = FALSE, rem_regout = FALSE, 
                 m = 100, nskip = 3, verbose = TRUE, seed = 123)
pred_surv <- pred_S3VS_SURV(y = y_surv, X = X[,res_surv$selected], 
                            surv_model = "COX", method = "COXGLMNET")
pred_surv
# Simulate survival data (Cox)
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
eta <- X[,1] + 0.5 * X[,2]
base_rate <- 0.05
T_event <- rexp(n, rate = base_rate * exp(eta))
C <- rexp(n, rate = 0.03)
time <- pmin(T_event, C)
status <- as.integer(T_event <= C)
y_surv <- list(time = time, status = status)
# Run S3VS for linear models
res_surv <- S3VS(y = y_surv, X = X, family = "survival", 
                 surv_model = "COX", vsel_method = "COXGLMNET",
                 method_xy = "topk", param_xy = list(k = 1),
                 method_xx = "topk", param_xx = list(k = 3),
                 method_sel = "conservative", method_rem = "conservative_begin",
                 sel_regout = FALSE, rem_regout = FALSE, 
                 m = 100, nskip = 3, verbose = TRUE, seed = 123)
pred_surv <- pred_S3VS_SURV(y = y_surv, X = X[,res_surv$selected], 
                            surv_model = "COX", method = "COXGLMNET")
pred_surv

Aggregate Not-Selected Predictors for Removal Across Multiple Leading Sets

Description

remove_vars combines lists of predictors that were not selected from multiple leading sets into a single set to remove, using either a liberal (union) rule or a conservative (progressive intersection) rule.

Usage

remove_vars(listnotselect, 
  method = c("conservative_begin", "conservative_end", "liberal"))
remove_vars(listnotselect, 
  method = c("conservative_begin", "conservative_end", "liberal"))

Arguments

listnotselect

A list of vectors, each containing names of the predictors not selected in the corresponding leading set.

method

Aggregation rule; one of "conservative_begin", "conservative_end", or "liberal". Referring to the sets (vectors included in listselect) of not-selected predictors:

"liberal": Returns the union of all sets: unique(unlist(listnotselect)).
"conservative_begin": Returns the last non-empty intersection when intersecting the first $i=1,2,\dots$ sets in order (note that, the first set is assumed to be non-empty, because that will automatically be true if remove_vars is being called by S3VS or its family-specific engines). The procedure stops once the running intersection becomes empty and returns the previous (last non-empty) intersection. Order of listnotselect matters.
"conservative_end": Returns the last non-empty intersection when intersecting the last $i=1,2,\dots$ sets in order (if the last set is empty, the function finds the first non-empty set from the end, and then, starts the intersection process from that set). The procedure stops once the running intersection becomes empty and returns the previous (last non-empty) intersection. Order of listnotselect matters.

Details

The liberal rule favors inclusiveness (drop all predictors that were not selected in an iteration), whereas the conservative rule favors stability across earlier/latter leading sets (drop only predictors consistently absent in earlier/latter leading sets).

Value

Vector with names of the predictors that are not selected till the current S3VS iteration and to be removed from all future iterations.

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

listselect <- list(
  c("V1","V2","V23"),
  c("V4","V2","V23"),
  c("V4","V5","V23")
)
remove_vars(listselect, method="liberal")
listselect <- list(
  c("V1","V2","V23"),
  c("V4","V2","V23"),
  c("V4","V5","V23")
)
remove_vars(listselect, method="liberal")

Structured Screen-and-Select Variable Selection in Linear, Generalized Linear, and Survival Models

Description

S3VS is the main function that performs variable selection based on the structured screen-and-select framework in linear, generalized linear, and survival models.

Usage

S3VS(
  y, 
  X, 
  family = c("normal", "binomial", "survival"), 
  cor_xy = NULL, 
  surv_model = c("COX", "AFT"), 
  method_xy = c("topk", "fixedthresh", "percthresh"), 
  param_xy, 
  method_xx = c("topk", "fixedthresh", "percthresh"), 
  param_xx, 
  vsel_method = NULL, 
  alpha = 0.5,
  method_sel = c("conservative", "liberal"), 
  method_rem = c("conservative_begin", "conservative_end", "liberal"), 
  sel_regout = FALSE, 
  rem_regout = FALSE, 
  update_y_thresh = 0.5, 
  m = 100, 
  nskip = 3, 
  verbose = FALSE, 
  seed = NULL, 
  parallel = FALSE
)
S3VS(
  y, 
  X, 
  family = c("normal", "binomial", "survival"), 
  cor_xy = NULL, 
  surv_model = c("COX", "AFT"), 
  method_xy = c("topk", "fixedthresh", "percthresh"), 
  param_xy, 
  method_xx = c("topk", "fixedthresh", "percthresh"), 
  param_xx, 
  vsel_method = NULL, 
  alpha = 0.5,
  method_sel = c("conservative", "liberal"), 
  method_rem = c("conservative_begin", "conservative_end", "liberal"), 
  sel_regout = FALSE, 
  rem_regout = FALSE, 
  update_y_thresh = 0.5, 
  m = 100, 
  nskip = 3, 
  verbose = FALSE, 
  seed = NULL, 
  parallel = FALSE
)

Arguments

y

X

Design matrix of predictors. Can be a base matrix or something as.matrix() can coerce. No missing values are allowed.

family

Model family; one of c("normal","binomial","survival"). Determines which engine is called (S3VS_LM, S3VS_GLM, or S3VS_SURV).

cor_xy

Optional numeric vector of precomputed marginal correlations between y and each column of X. Used only when family="normal" to speed up or reproduce screening by $|cor(y, X_j)|$ . If NULL, correlations are computed internally.

surv_model

Character string specifying the survival model (for family="survival" only). Must be explicitly provided; there is no default. Values are "Cox" for proportional hazards models, "AFT" for accelerated failure time models.

method_xy

Rule for screening some predictors as "leading variables" based on their association with the response; one of c("topk", "fixedthresh", "percthresh"). The association measure depends on family (e.g., correlation for "normal", eta-squared for "binomial", or marginal utility for "survival").

"topk" keeps the predictors with the largest $k$ association values; "fixedthresh" keeps predictors whose association is greater than or equal to a specified threshold; "percthresh" keeps predictors whose association is within a given percentage of the best.

param_xy

Tuning parameter for method_xy. If "topk", supply a list with an integer k (keep the top $k$ ). If "fixedthresh", supply a list with a numeric threshold thresh (keep predictors with association $\ge$ threshold). If "percthresh", supply a list with a numeric percentage thresh in $(0,100]$ (keep predictors with association $\ge$ that percent of the highest association).

method_xx

param_xx

Tuning parameter for method_xx; same interpretation as param_xy but applied to inter-predictor association (absolute value of the correlation coefficient).

vsel_method

Character string specifying the variable selection method to be used within each leading set. Available options depend on the model type:

For linear models (S3VS_LM) and generalized linear models (S3VS_GLM): "NLP", "LASSO", "ENET", "SCAD", "MCP".
For survival models (S3VS_SURV): "LASSO", "ENET" for surv_model=COX and "AFTGEE", "BRIDGE", "PVAFT" for surv_model=AFT.

alpha

Only used when vsel_method == "ENET". Elastic net mixing parameter, with $\alpha \in (0,1)$ .

method_sel

Policy for aggregating predictors selected across leading sets in an iteration; one of c("conservative","liberal"). "conservative" selects the smallest admissible set of predictors by intersecting the selected sets of predictors across leading sets, beginning with all and gradually reducing from the end until a non-empty intersection is found; this ensures only predictors consistently selected across leading sets are retained. "liberal" selects the largest admissible set of predictors by taking the union of all selected sets of predictors, so any predictor chosen in at least one leading set is included. If no predictor is selected from the first leading set, the iteration does not contribute to final selection and exclusion rules (method_rem) are applied instead.

method_rem

Policy for excluding predictors when no selections are made in an iteration; one of c("conservative_begin","conservative_end","liberal"). "conservative_begin" excludes the smallest admissible set of predictors by intersecting the non-selected sets of predictors starting from the first leading set; "conservative_end" does the same but begins from the last leading set and moves backward; "liberal" excludes the largest admissible set of predictors by taking the union of all non-selected sets of predictor. Predictors excluded under this rule are removed from subsequent iterations.

sel_regout

Logical (GLM only). If TRUE, when predictors are selected in an iteration, the working response y is updated using the selected predictors via update_y_GLM. Ignored for other families.

rem_regout

Logical (for LM and GLM only). If TRUE, when no predictors are selected in an iteration and some are removed instead, the working response y is updated using the removed predictors via update_y_LM or update_y_GLM. Ignored for other families.

update_y_thresh

Numeric scalar threshold controlling how the working response y is updated in GLMs when sel_regout=TRUE or rem_regout=TRUE. When $|y - fitted\_y| > update\_y\_thresh$ , y is kept, else y replaced by the rounded value of fitted_y, where fitted_y is the fitted probability from the logistic model. The default value is 0.5. Ignored for other families.

m

Integer. Maximum number of S3VS iterations to perform. Defaults to 100.

nskip

Integer. Maximum number of iterations in which no new predictors are selected before the algorithm stops. Defaults to 3.

verbose

Logical. If TRUE, prints detailed progress information at each iteration (e.g., iteration number, predictors selected or removed). Defaults to FALSE.

seed

If supplied, sets the random seed via set.seed() to ensure reproducibility of stochastic components. If NULL, no seed is set.

parallel

Logical. If TRUE, attempts to perform some computations in parallel mode in binomial and survival families, which is strongly recommended for faster execution. Defaults to FALSE.

Details

Model

For a continuous response, S3VS considers the linear model (LM)

$\boldsymbol{y} = \boldsymbol{X}\boldsymbol{\beta} + \boldsymbol{\epsilon}$

For a binary response, S3VS considers the generalized linear model (GLM)

$g\!\left( E\!\left( \boldsymbol{y} \mid \boldsymbol{X} \right) \right) = \boldsymbol{X}\boldsymbol{\beta}$

For a survival type response, S3VS considers two choices of models–the Cox model

$\lambda(t\mid \boldsymbol{x}_i) = \lambda_0(t) \exp(\boldsymbol{x}_i^T \boldsymbol{\beta})$

and the AFT model

$\log(\boldsymbol{T}) = \boldsymbol{X}\boldsymbol{\beta} + \boldsymbol{\epsilon}$

S3VS algorithm

The general form of the S3VS algorithm consists of the following steps, repeated iteratively until convergence:

Determination of leading variables: 'Leading variables' are determined based on the association of the predictors with the response, following one of three rules. The rule is fixed by the arguments method_xy and param_xy.
Determination of leading sets: For each leading variable, a group of related predictors, called the 'leading set', is determined based on the association of all candidate predictors with the leading variable, following one of three rules. The rule is fixed by the arguments method_xx and param_xx.
Variable selection: Within each leading set, small to moderate-dimensional variable selection is performed using a method fixed by vsel_method.
Aggregation of selected/not-selected variables: Variables selected/not-selected in different leading sets are aggregated using several possible rules, fixed by method_sel and method_rem.
Updation of response and/or set of covariates: At the end of each iteration, the response and predictors may be chosen to be updated or not through argumentsm sel_regout, rem_regout, and update_y_thresh.

The convergence criterion is determined by the arguments m and nkip jointly. For ore details of the individual steps, see the manual of the functions linked below.

Value

A list with the following components:

selected

A character vector of predictor names that were selected across all iterations.

selected_iterwise

A list recording the predictors selected at each iteration, in the order they were considered.

runtime

Runtime in seconds.

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

### [1] For linear model
# Simulate continuous data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
y <- X[,1] + 0.5 * X[,2] + rnorm(n)
# Run S3VS for LM
res_lm <- S3VS(y = y, X = X, family = "normal",
               method_xy = "topk", param_xy = list(k=1),
               method_xx = "topk", param_xx = list(k=3),
               vsel_method = "LASSO", method_sel = "conservative", 
               method_rem = "conservative_begin", rem_regout = FALSE, 
               m = 100, nskip = 3, verbose = TRUE, seed = 123)
# View selected predictors
res_lm$selected

### [2] For generalized linear model
# Simulate binary data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
eta <- X[,1] + 0.5 * X[,2]
prob <- 1 / (1 + exp(-eta))
y <- rbinom(n, size = 1, prob = prob)
# Run S3VS for for GLM (logistic)
res_glm <- S3VS(y = y, X = X, family = "binomial",
                method_xy = "topk", param_xy = list(k = 1),
                method_xx = "topk", param_xx = list(k = 3),
                vsel_method = "LASSO", 
                method_sel = "conservative", method_rem = "conservative_begin", 
                sel_regout = FALSE, rem_regout = FALSE,
                m = 100, nskip = 3, verbose = TRUE, seed = 123)
# View selected predictors
res_glm$selected

### [3] For survival model
# Simulate survival data (Cox)
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
eta <- X[,1] + 0.5 * X[,2]
base_rate <- 0.05
T_event <- rexp(n, rate = base_rate * exp(eta))
C <- rexp(n, rate = 0.03)
time <- pmin(T_event, C)
status <- as.integer(T_event <= C)
y_surv <- list(time = time, status = status)
# Run S3VS for linear models
res_surv <- S3VS(y = y_surv, X = X, family = "survival", 
                 surv_model = "COX", 
                 method_xy = "topk", param_xy = list(k = 1),
                 method_xx = "topk", param_xx = list(k = 3),
                 vsel_method = "COXGLMNET",
                 method_sel = "conservative", method_rem = "conservative_begin",
                 sel_regout = FALSE, rem_regout = FALSE, 
                 m = 100, nskip = 3, verbose = TRUE, seed = 123)
# View selected predictors
res_surv$selected
### [1] For linear model
# Simulate continuous data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
y <- X[,1] + 0.5 * X[,2] + rnorm(n)
# Run S3VS for LM
res_lm <- S3VS(y = y, X = X, family = "normal",
               method_xy = "topk", param_xy = list(k=1),
               method_xx = "topk", param_xx = list(k=3),
               vsel_method = "LASSO", method_sel = "conservative", 
               method_rem = "conservative_begin", rem_regout = FALSE, 
               m = 100, nskip = 3, verbose = TRUE, seed = 123)
# View selected predictors
res_lm$selected

### [2] For generalized linear model
# Simulate binary data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
eta <- X[,1] + 0.5 * X[,2]
prob <- 1 / (1 + exp(-eta))
y <- rbinom(n, size = 1, prob = prob)
# Run S3VS for for GLM (logistic)
res_glm <- S3VS(y = y, X = X, family = "binomial",
                method_xy = "topk", param_xy = list(k = 1),
                method_xx = "topk", param_xx = list(k = 3),
                vsel_method = "LASSO", 
                method_sel = "conservative", method_rem = "conservative_begin", 
                sel_regout = FALSE, rem_regout = FALSE,
                m = 100, nskip = 3, verbose = TRUE, seed = 123)
# View selected predictors
res_glm$selected

### [3] For survival model
# Simulate survival data (Cox)
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
eta <- X[,1] + 0.5 * X[,2]
base_rate <- 0.05
T_event <- rexp(n, rate = base_rate * exp(eta))
C <- rexp(n, rate = 0.03)
time <- pmin(T_event, C)
status <- as.integer(T_event <= C)
y_surv <- list(time = time, status = status)
# Run S3VS for linear models
res_surv <- S3VS(y = y_surv, X = X, family = "survival", 
                 surv_model = "COX", 
                 method_xy = "topk", param_xy = list(k = 1),
                 method_xx = "topk", param_xx = list(k = 3),
                 vsel_method = "COXGLMNET",
                 method_sel = "conservative", method_rem = "conservative_begin",
                 sel_regout = FALSE, rem_regout = FALSE, 
                 m = 100, nskip = 3, verbose = TRUE, seed = 123)
# View selected predictors
res_surv$selected

Structured Screen-and-Select Variable Selection in Generalized Linear Models

Description

S3VS_GLM performs variable selection based on the structured screen-and-select framework in generalized linear models.

Usage

S3VS_GLM(y, X, 
  method_xy = c("topk", "fixedetasqthresh", "percetasqthresh"), param_xy, 
  method_xx = c("topk", "fixedcorthresh", "perccorthresh"), param_xx, 
  vsel_method = c("NLP", "LASSO", "ENET", "SCAD", "MCP"), 
  alpha = 0.5,
  method_sel = c("conservative", "liberal"), 
  method_rem = c("conservative_begin", "conservative_end", "liberal"), 
  sel_regout = FALSE, rem_regout = FALSE, update_y_thresh = NULL, 
  m = 100, nskip = 3, verbose = FALSE, seed = NULL, parallel = FALSE)
S3VS_GLM(y, X, 
  method_xy = c("topk", "fixedetasqthresh", "percetasqthresh"), param_xy, 
  method_xx = c("topk", "fixedcorthresh", "perccorthresh"), param_xx, 
  vsel_method = c("NLP", "LASSO", "ENET", "SCAD", "MCP"), 
  alpha = 0.5,
  method_sel = c("conservative", "liberal"), 
  method_rem = c("conservative_begin", "conservative_end", "liberal"), 
  sel_regout = FALSE, rem_regout = FALSE, update_y_thresh = NULL, 
  m = 100, nskip = 3, verbose = FALSE, seed = NULL, parallel = FALSE)

Arguments

y

Response. A numeric/integer/logical vector with values in {0,1}.

X

Design matrix of predictors. Can be a base matrix or something as.matrix() can coerce. No missing values are allowed.

method_xy

Rule for screening some predictors as "leading variables" based on their association with the response; one of c("topk", "fixedthresh", "percthresh"). The association measure is eta-squared.

param_xy

method_xx

param_xx

Tuning parameter for method_xx; same interpretation as param_xy but applied to inter-predictor association (absolute value of the correlation coefficient).

vsel_method

Character string specifying the variable selection method to be used within each leading set. Available options are "NLP", "LASSO", "ENET", "SCAD", "MCP".

alpha

Only used when vsel_method == "ENET". Elastic net mixing parameter, with $\alpha \in (0,1)$ .

method_sel

method_rem

sel_regout

Logical. If TRUE, when predictors are selected in an iteration, the working response y is updated using the selected predictors via update_y_GLM.

rem_regout

Logical. If TRUE, when no predictors are selected in an iteration and some are removed instead, the working response y is updated using the removed predictors via update_y_GLM.

update_y_thresh

Numeric scalar threshold controlling how the working response y is updated when sel_regout=TRUE or rem_regout=TRUE. When $|y - fitted\_y| > update\_y\_thresh$ , y is kept, else y replaced by the rounded value of fitted_y, where fitted_y is the fitted probability from the logistic model. The default value is 0.5.

m

Integer. Maximum number of S3VS iterations to perform. Defaults to 100.

nskip

Integer. Maximum number of iterations in which no new predictors are selected before the algorithm stops. Defaults to 3.

verbose

Logical. If TRUE, prints detailed progress information at each iteration (e.g., iteration number, predictors selected or removed). Defaults to FALSE.

seed

If supplied, sets the random seed via set.seed() to ensure reproducibility of stochastic components. If NULL, no seed is set.

parallel

Logical. If TRUE, attempts to perform some computations in parallel mode, which is strongly recommended for faster execution. Defaults to FALSE.

Details

For a binary response, S3VS considers the generalized linear model (GLM)

$g\!\left( E\!\left( \boldsymbol{y} \mid \boldsymbol{X} \right) \right) = \boldsymbol{X}\boldsymbol{\beta}$

For the S3VS algorithm, see the manual of the top-level function S3VS.

Value

A list with the following components:

selected

A character vector of predictor names that were selected across all iterations.

selected_iterwise

A list recording the predictors selected at each iteration, in the order they were considered.

runtime

Runtime in seconds.

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

# Simulate binary data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
eta <- X[,1] + 0.5 * X[,2]
prob <- 1 / (1 + exp(-eta))
y <- rbinom(n, size = 1, prob = prob)
# Run S3VS for for GLM (logistic)
res_glm <- S3VS_GLM(y = y, X = X,
                method_xy = "topk", param_xy = list(k = 1),
                method_xx = "topk", param_xx = list(k = 3),
                vsel_method = "LASSO", 
                method_sel = "conservative", method_rem = "conservative_begin", 
                sel_regout = FALSE, rem_regout = FALSE,
                m = 100, nskip = 3, verbose = TRUE, seed = 123)
# View selected predictors
res_glm$selected
# Simulate binary data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
eta <- X[,1] + 0.5 * X[,2]
prob <- 1 / (1 + exp(-eta))
y <- rbinom(n, size = 1, prob = prob)
# Run S3VS for for GLM (logistic)
res_glm <- S3VS_GLM(y = y, X = X,
                method_xy = "topk", param_xy = list(k = 1),
                method_xx = "topk", param_xx = list(k = 3),
                vsel_method = "LASSO", 
                method_sel = "conservative", method_rem = "conservative_begin", 
                sel_regout = FALSE, rem_regout = FALSE,
                m = 100, nskip = 3, verbose = TRUE, seed = 123)
# View selected predictors
res_glm$selected

Structured Screen-and-Select Variable Selection in Linear Models

Description

S3VS_LM performs variable selection based on the structured screen-and-select framework in linear models.

Usage

S3VS_LM(y, X, cor_xy = NULL, 
  method_xy = c("topk", "fixedcorthresh", "perccorthresh"), param_xy, 
  method_xx = c("topk", "fixedcorthresh", "perccorthresh"), param_xx, 
  vsel_method = c("NLP", "LASSO", "ENET", "SCAD", "MCP"), 
  alpha = 0.5,
  method_sel = c("conservative", "liberal"), 
  method_rem = c("conservative_begin", "conservative_end", "liberal"), 
  rem_regout = FALSE, 
  m = 100, nskip = 3, verbose = FALSE, seed = NULL)
S3VS_LM(y, X, cor_xy = NULL, 
  method_xy = c("topk", "fixedcorthresh", "perccorthresh"), param_xy, 
  method_xx = c("topk", "fixedcorthresh", "perccorthresh"), param_xx, 
  vsel_method = c("NLP", "LASSO", "ENET", "SCAD", "MCP"), 
  alpha = 0.5,
  method_sel = c("conservative", "liberal"), 
  method_rem = c("conservative_begin", "conservative_end", "liberal"), 
  rem_regout = FALSE, 
  m = 100, nskip = 3, verbose = FALSE, seed = NULL)

Arguments

y

Response. A numeric vector.

X

Design matrix of predictors. Can be a base matrix or something as.matrix() can coerce. No missing values are allowed.

cor_xy

Optional numeric vector of precomputed marginal correlations between y and each column of X. Used to speed up or reproduce screening by $|cor(y, X_j)|$ . If NULL, correlations are computed internally.

method_xy

Rule for screening some predictors as 'leading variables' based on their association with the response; one of c("topk", "fixedthresh", "percthresh"). The association measure is correlation.

param_xy

method_xx

param_xx

Tuning parameter for method_xx; same interpretation as param_xy but applied to inter-predictor association (absolute value of the correlation coefficient).

vsel_method

Character string specifying the variable selection method to be used within each leading set. Available options are "NLP", "LASSO", "ENET", "SCAD", "MCP".

alpha

Only used when vsel_method == "ENET". Elastic net mixing parameter, with $\alpha \in (0,1)$ .

method_sel

method_rem

rem_regout

Logical. If TRUE, when no predictors are selected in an iteration and some are removed instead, the working response y is updated using the removed predictors via update_y_LM.

m

Integer. Maximum number of S3VS iterations to perform. Defaults to 100.

nskip

Integer. Maximum number of iterations in which no new predictors are selected before the algorithm stops. Defaults to 3.

verbose

Logical. If TRUE, prints detailed progress information at each iteration (e.g., iteration number, predictors selected or removed). Defaults to FALSE.

seed

If supplied, sets the random seed via set.seed() to ensure reproducibility of stochastic components. If NULL, no seed is set.

Details

For a continuous response, S3VS considers the linear model (LM)

$\boldsymbol{y} = \boldsymbol{X}\boldsymbol{\beta} + \boldsymbol{\epsilon}$

For the S3VS algorithm, see the manual of the top-level function S3VS.

Value

A list with the following components:

selected

A character vector of predictor names that were selected across all iterations.

selected_iterwise

A list recording the predictors selected at each iteration, in the order they were considered.

runtime

Runtime in seconds.

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

# Simulate continuous data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
y <- X[,1] + 0.5 * X[,2] + rnorm(n)
# Run S3VS for LM
res_lm <- S3VS_LM(y = y, X = X,
               method_xy = "topk", param_xy = list(k=1),
               method_xx = "topk", param_xx = list(k=3),
               vsel_method = "LASSO", method_sel = "conservative", 
               method_rem = "conservative_begin", rem_regout = FALSE, 
               m = 100, nskip = 3, verbose = TRUE, seed = 123)
# View selected predictor
res_lm$selected
# Simulate continuous data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
y <- X[,1] + 0.5 * X[,2] + rnorm(n)
# Run S3VS for LM
res_lm <- S3VS_LM(y = y, X = X,
               method_xy = "topk", param_xy = list(k=1),
               method_xx = "topk", param_xx = list(k=3),
               vsel_method = "LASSO", method_sel = "conservative", 
               method_rem = "conservative_begin", rem_regout = FALSE, 
               m = 100, nskip = 3, verbose = TRUE, seed = 123)
# View selected predictor
res_lm$selected

Structured Screen-and-Select Variable Selection in Survival Models

Description

S3VS_SURV performs variable selection based on the structured screen-and-select framework in survival models.

Usage

S3VS_SURV(y, X, surv_model = c("COX", "AFT"), 
  method_xy = c("topk", "fixedmuthresh", "percmuthresh"), param_xy, 
  method_xx = c("topk", "fixedcorthresh", "perccorthresh"), param_xx, 
  vsel_method = c("LASSO", "ENET", "AFTGEE", "BRIDGE", "PVAFT"), 
  alpha = 0.5,
  method_sel = c("conservative", "liberal"), 
  method_rem = c("conservative_begin", "conservative_end", "liberal"), 
  m = 100, nskip = 3, verbose = FALSE, seed = NULL, parallel = FALSE)
S3VS_SURV(y, X, surv_model = c("COX", "AFT"), 
  method_xy = c("topk", "fixedmuthresh", "percmuthresh"), param_xy, 
  method_xx = c("topk", "fixedcorthresh", "perccorthresh"), param_xx, 
  vsel_method = c("LASSO", "ENET", "AFTGEE", "BRIDGE", "PVAFT"), 
  alpha = 0.5,
  method_sel = c("conservative", "liberal"), 
  method_rem = c("conservative_begin", "conservative_end", "liberal"), 
  m = 100, nskip = 3, verbose = FALSE, seed = NULL, parallel = FALSE)

Arguments

y

Response. A list with components time and status (1 = event, 0 = censored).

X

Design matrix of predictors. Can be a base matrix or something as.matrix() can coerce. No missing values are allowed.

surv_model

Character string specifying the survival model. Must be explicitly provided; there is no default. Values are "Cox" for proportional hazards models, "AFT" for accelerated failure time models.

method_xy

Rule for screening some predictors as "leading variables" based on their association with the response; one of c("topk", "fixedthresh", "percthresh"). The association measure is marginal utility.

param_xy

method_xx

param_xx

Tuning parameter for method_xx; same interpretation as param_xy but applied to inter-predictor association (absolute value of the correlation coefficient).

vsel_method

Character string specifying the variable selection method to be used within each leading set. Available options are "LASSO", "ENET" for surv_model=COX and "AFTGEE", "BRIDGE", "PVAFT" for surv_model=AFT.

alpha

Only used when vsel_method == "ENET". Elastic net mixing parameter, with $\alpha \in (0,1)$ .

method_sel

method_rem

m

Integer. Maximum number of S3VS iterations to perform. Defaults to 100.

nskip

Integer. Maximum number of iterations in which no new predictors are selected before the algorithm stops. Defaults to 3.

verbose

Logical. If TRUE, prints detailed progress information at each iteration (e.g., iteration number, predictors selected or removed). Defaults to FALSE.

seed

If supplied, sets the random seed via set.seed() to ensure reproducibility of stochastic components. If NULL, no seed is set.

parallel

Logical. If TRUE, attempts to perform some computations in parallel mode, which is strongly recommended for faster execution. Defaults to FALSE.

Details

For a survival type response, S3VS considers two choices of models–the Cox model

$\lambda(t\mid \boldsymbol{x}_i) = \lambda_0(t) \exp(\boldsymbol{x}_i^T \boldsymbol{\beta})$

and the AFT model

$\log(\boldsymbol{T}) = \boldsymbol{X}\boldsymbol{\beta} + \boldsymbol{\epsilon}$

For the S3VS algorithm, see the manual of the top-level function S3VS.

Value

A list with the following components:

selected

A character vector of predictor names that were selected across all iterations.

selected_iterwise

A list recording the predictors selected at each iteration, in the order they were considered.

runtime

Runtime in seconds.

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

# Simulate survival data (Cox)
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
eta <- X[,1] + 0.5 * X[,2]
base_rate <- 0.05
T_event <- rexp(n, rate = base_rate * exp(eta))
C <- rexp(n, rate = 0.03)
time <- pmin(T_event, C)
status <- as.integer(T_event <= C)
y_surv <- list(time = time, status = status)
# Run S3VS for linear models
res_surv <- S3VS(y = y_surv, X = X, family = "survival", 
                 surv_model = "COX", 
                 method_xy = "topk", param_xy = list(k = 1),
                 method_xx = "topk", param_xx = list(k = 3),
                 vsel_method = "COXGLMNET",
                 method_sel = "conservative", method_rem = "conservative_begin",
                 sel_regout = FALSE, rem_regout = FALSE, 
                 m = 100, nskip = 3, verbose = TRUE, seed = 123)
# View selected predictors
res_surv$selected
# Simulate survival data (Cox)
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
eta <- X[,1] + 0.5 * X[,2]
base_rate <- 0.05
T_event <- rexp(n, rate = base_rate * exp(eta))
C <- rexp(n, rate = 0.03)
time <- pmin(T_event, C)
status <- as.integer(T_event <= C)
y_surv <- list(time = time, status = status)
# Run S3VS for linear models
res_surv <- S3VS(y = y_surv, X = X, family = "survival", 
                 surv_model = "COX", 
                 method_xy = "topk", param_xy = list(k = 1),
                 method_xx = "topk", param_xx = list(k = 3),
                 vsel_method = "COXGLMNET",
                 method_sel = "conservative", method_rem = "conservative_begin",
                 sel_regout = FALSE, rem_regout = FALSE, 
                 m = 100, nskip = 3, verbose = TRUE, seed = 123)
# View selected predictors
res_surv$selected

Aggregate Selected Predictors Across Multiple Leading Sets

Description

select_vars combines variable selections obtained from multiple leading sets into a single set, using either a liberal (union) or conservative (progressive intersection) rule.

Usage

select_vars(listselect, method = c("conservative", "liberal"))
select_vars(listselect, method = c("conservative", "liberal"))

Arguments

listselect

A list of vectors, each containing names of the predictors selected in the corresponding leading set.

method

Aggregation rule. One of "conservative" or "liberal" (partial matching allowed). Referring to the sets (vectors included in listselect) of selected predictors:

"liberal": Returns the union of all sets: unique(unlist(listselect)).
"conservative": Returns the last non-empty intersection when intersecting the first $i=1,2,\dots$ sets in order. The procedure stops once the running intersection becomes empty and returns the previous (last non-empty) intersection. If the first set is empty, returns character(0). Order of vectors in listselect matters for this method.

Details

The liberal rule favors inclusiveness, while the conservative rule favors stability.

Value

Vector with names of the retained predictors (considered selected in the current iteration of S3VS); if no predictors are retained, character(0)).

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

listselect <- list(
  c("V1","V2","V23"),
  c("V4","V2","V23"),
  c("V4","V5","V23")
)
select_vars(listselect, method="conservative")
listselect <- list(
  c("V1","V2","V23"),
  c("V4","V2","V23"),
  c("V4","V5","V23")
)
select_vars(listselect, method="conservative")

Update Response Accounting for Selected Predictors

Description

update_y updates the response accounting for the selected predictors in linear models, and selected or removed predictors in generalized linear models.

Usage

update_y(y, X, family, vars, update_y_thresh = NULL)
update_y(y, X, family, vars, update_y_thresh = NULL)

Arguments

y

Response. If family = "normal", a numeric vector. If family = "binomial", a numeric/integer/logical vector with values in {0,1}.

family

Model family; one of c("normal","binomial"). Determines which engine is called (update_y_LM, update_y_GLM).

X

Predictor matrix. Can be a base matrix or something as.matrix() can coerce. No missing values are allowed.

vars

Character vector containing the names of predictors that need to be accounted for. They must appear in X.

update_y_thresh

Numeric scalar threshold used only family = "binomial". When $|y - fitted\_y| > update\_y\_thresh$ , y is kept, else y replaced by the rounded value of $fitted_y$ , which is the fitted probability from the logistic model. The default value is 0.5.

Value

Returns the updated response vector.

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

# Simulate continuous data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
y <- X[,1] + 0.5 * X[,2] + rnorm(n)
update_y(y = y, X = X, family = "normal", vars = c("V1","V4"))
# Simulate continuous data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
y <- X[,1] + 0.5 * X[,2] + rnorm(n)
update_y(y = y, X = X, family = "normal", vars = c("V1","V4"))

Update Response Accounting for Selected Predictors in Generalized Linear Models

Description

update_y_LM updates the response accounting for the selected predictors in generalized linear models.

Usage

update_y_GLM(y, X, vars, update_y_thresh)
update_y_GLM(y, X, vars, update_y_thresh)

Arguments

y

Response. A numeric/integer/logical vector with values in {0,1}.

X

Predictor matrix. Can be a base matrix or something as.matrix() can coerce. No missing values are allowed.

vars

Character vector containing the names of predictors that need to be accounted for. They must appear in X.

update_y_thresh

Numeric scalar threshold. When $|y - fitted\_y| > update\_y\_thresh$ , y is kept, else y replaced by the rounded value of fitted_y, which is the fitted probability from the logistic model. The default value is 0.5.

Value

Returns the updated (binary) response vector.

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

# Simulate binary data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
eta <- X[,1] + 0.5 * X[,2]
prob <- 1 / (1 + exp(-eta))
y <- rbinom(n, size = 1, prob = prob)
update_y(family = "binomial", y = y, X = X, vars = c("V1","V4"), update_y_thresh = 0.8)
# Simulate binary data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
eta <- X[,1] + 0.5 * X[,2]
prob <- 1 / (1 + exp(-eta))
y <- rbinom(n, size = 1, prob = prob)
update_y(family = "binomial", y = y, X = X, vars = c("V1","V4"), update_y_thresh = 0.8)

Update Response Accounting for Selected Predictors in Linear Models

Description

update_y_LM updates the response accounting for the selected predictors in linear models.

Usage

update_y_LM(y, X, vars)
update_y_LM(y, X, vars)

Arguments

y

Response. A numeric vector of length $n$ .

X

Predictor matrix. Can be a base matrix or something as.matrix() can coerce. No missing values are allowed.

vars

Character vector containing the names of predictors that need to be accounted for. They must appear in X.

Value

Returns the updated response vector.

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

# Simulate continuous data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
y <- X[,1] + 0.5 * X[,2] + rnorm(n)
update_y(family = "normal", y = y, X = X, vars = c("V1","V4"))
# Simulate continuous data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
y <- X[,1] + 0.5 * X[,2] + rnorm(n)
update_y(family = "normal", y = y, X = X, vars = c("V1","V4"))

Variable Selection in Leading Sets under the S3VS Framework

Description

VS_method applies the chosen variable-selection algorithm to each leading set produced by S3VS at every iteration.

Usage

VS_method(y, X, family, surv_model = NULL, vsel_method, alpha = 0.5,
          p_thresh = 0.1, gamma = 0.9, verbose = FALSE)
VS_method(y, X, family, surv_model = NULL, vsel_method, alpha = 0.5,
          p_thresh = 0.1, gamma = 0.9, verbose = FALSE)

Arguments

y

X

Predictor matrix. Can be a base matrix or something as.matrix() can coerce. No missing values are allowed.

family

Model family; one of c("normal","binomial","survival"). Determines which engine is called (VS_method_LM, VS_method_GLM, or VS_method_SURV).

surv_model

vsel_method

Character string indicating the variable-selection engine used inside S3VS at each iteration. Allowed values depend on family: for family = "normal"(S3VS_LM) and "binomial" (S3VS_GLM), available options are "NLP", "ENET", "LASSO", "SCAD", "MCP"; for family = "survival", available options are "LASSO", "ENET" for surv_model = "COX" and "AFTGEE", "BRIDGE", "PVAFT" for surv_model = "AFT". See Details for more information.

alpha

Only used when vsel_method == "ENET". Elastic net mixing parameter, with $\alpha \in (0,1)$ .

p_thresh

Only used for surv_model = "AFT" with vsel_method = "AFTGEE". p-value threshold for variable selection.

gamma

Only used for surv_model = "AFT" with vsel_method = "BRIDGE". Numeric scalar (default 0.5) giving the exponent in the bridge penalty. Must be $> 0$ . See details.

verbose

If TRUE, some information is printed. FALSE, by default.

Details

Details to come...

Value

A list containing:

sel

Character vector with names of the selected predictors.

nosel

Character vector with names of the predictors not selected.

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

# Simulate continuous data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
y <- X[,1] + 0.5 * X[,2] + rnorm(n)
# Run VS_method
VS_method(y, X, family = "normal", vsel_method = "NLP", verbose = FALSE)
# Simulate continuous data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
y <- X[,1] + 0.5 * X[,2] + rnorm(n)
# Run VS_method
VS_method(y, X, family = "normal", vsel_method = "NLP", verbose = FALSE)

Variable Selection in Leading Sets for Generalized Linear Models under the S3VS Framework

Description

VS_method applies the chosen variable-selection algorithm for generalized linear models to each leading set produced by S3VS at every iteration.

Usage

VS_method_GLM(y, X, vsel_method, alpha = 0.5, verbose = FALSE, 
              parallel = FALSE, ncores = NULL)
VS_method_GLM(y, X, vsel_method, alpha = 0.5, verbose = FALSE, 
              parallel = FALSE, ncores = NULL)

Arguments

y

Response. A numeric/integer/logical vector with values in {0,1}.

X

Predictor matrix. Can be a base matrix or something as.matrix() can coerce. No missing values are allowed.

vsel_method

Character string indicating the variable-selection engine used at each iteration. Available options are "NLP" and "LASSO". See Details for more information.

alpha

Only used when vsel_method == "ENET". Elastic net mixing parameter, with $\alpha \in (0,1)$ .

verbose

If TRUE, some information is printed. FALSE, by default.

parallel

Logical. If TRUE, cross-validation steps in penalized regression methods (e.g., LASSO or elastic net via glmnet) are executed in parallel using the doParallel backend. Defaults to FALSE. The package doParallel must be installed when parallel = TRUE.

ncores

Integer; number of CPU cores to use when parallel = TRUE. If NULL (default), the number of cores is set to one fewer than the total number detected on the system. Ignored when parallel = FALSE.

Value

A list containing:

sel

Character vector with names of the selected predictors.

nosel

Character vector with names of the predictors not selected.

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

# Simulate binary data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
eta <- X[,1] + 0.5 * X[,2]
prob <- 1 / (1 + exp(-eta))
y <- rbinom(n, size = 1, prob = prob)
# Run VS_method
VS_method_GLM(y, X, vsel_method = "LASSO", verbose = FALSE)
# Simulate binary data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
eta <- X[,1] + 0.5 * X[,2]
prob <- 1 / (1 + exp(-eta))
y <- rbinom(n, size = 1, prob = prob)
# Run VS_method
VS_method_GLM(y, X, vsel_method = "LASSO", verbose = FALSE)

Variable Selection in Leading Sets for Linear Models under the S3VS Framework

Description

VS_method applies the chosen variable-selection algorithm for linear models to each leading set produced by S3VS at every iteration.

Usage

VS_method_LM(y, X, vsel_method, alpha = 0.5, verbose = FALSE)
VS_method_LM(y, X, vsel_method, alpha = 0.5, verbose = FALSE)

Arguments

y

Response. A numeric vector.

X

Predictor matrix. Can be a base matrix or something as.matrix() can coerce. No missing values are allowed.

vsel_method

Character string indicating the variable-selection engine used at each iteration. Available options are "NLP", "LASSO", "SCAD", "MCP". See Details for more information.

alpha

Only used when vsel_method == "ENET". Elastic net mixing parameter, with $\alpha \in (0,1)$ .

verbose

If TRUE, some information is printed. FALSE, by default.

Value

A list containing:

sel

Character vector with names of the selected predictors.

nosel

Character vector with names of the predictors not selected.

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

# Simulate continuous data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
y <- X[,1] + 0.5 * X[,2] + rnorm(n)
# Run VS_method
VS_method_LM(y, X, vsel_method = "NLP", verbose = FALSE)
# Simulate continuous data
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
y <- X[,1] + 0.5 * X[,2] + rnorm(n)
# Run VS_method
VS_method_LM(y, X, vsel_method = "NLP", verbose = FALSE)

Variable Selection in Leading Sets for Survival Models under the S3VS Framework

Description

VS_method applies the chosen variable-selection algorithm for survival models to each leading set produced by S3VS at every iteration.

Usage

VS_method_SURV(y, X, surv_model, vsel_method, alpha = 0.5,
  p_thresh = 0.1, gamma = 0.9, verbose = FALSE, ...)
VS_method_SURV(y, X, surv_model, vsel_method, alpha = 0.5,
  p_thresh = 0.1, gamma = 0.9, verbose = FALSE, ...)

Arguments

y

Response. A list with components time and status (1 = event, 0 = censored).

X

Predictor matrix. Can be a base matrix or something as.matrix() can coerce. No missing values are allowed.

surv_model

Character string specifying the survival model. Must be explicitly provided; there is no default. Values are "Cox" for proportional hazards models, "AFT" for accelerated failure time models.

vsel_method

Character string indicating the variable-selection engine used at each iteration. Available options are "COXGLMNET" for surv_model = "COX" and "AFTREG", "AFTGEE", "BRIDGE", and "PVAFT" for surv_model = "AFT". See Details for more information.

alpha

Only used when vsel_method == "ENET". Elastic net mixing parameter, with $\alpha \in (0,1)$ .

p_thresh

Only used with vsel_method = "AFTGEE". p-value threshold for variable selection.

gamma

Only used with vsel_method = "BRIDGE". Numeric scalar (default 0.5) giving the exponent in the bridge penalty. Must be $> 0$ . See details.

verbose

If TRUE, some information is printed. FALSE, by default.

...

Other arguments to be passed inside eha::aftreg() when surv_model = "AFT" and vsel_method = "AFTREG"

Value

A list containing:

sel

Character vector with names of the selected predictors.

nosel

Character vector with names of the predictors not selected.

Author(s)

Nilotpal Sanyal <[email protected]>, Padmore N. Prempeh <[email protected]>

Examples

# Simulate survival data (Cox)
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
eta <- X[,1] + 0.5 * X[,2]
base_rate <- 0.05
T_event <- rexp(n, rate = base_rate * exp(eta))
C <- rexp(n, rate = 0.03)
time <- pmin(T_event, C)
status <- as.integer(T_event <= C)
y_surv <- list(time = time, status = status)
# Run VS_method
VS_method_SURV(y_surv, X, surv_model = "COX", vsel_method = "COXGLMNET", verbose = FALSE)
# Simulate survival data (Cox)
set.seed(123)
n <- 100
p <- 150
X <- matrix(rnorm(n * p), n, p)
colnames(X) <- paste0("V", 1:p)
eta <- X[,1] + 0.5 * X[,2]
base_rate <- 0.05
T_event <- rexp(n, rate = base_rate * exp(eta))
C <- rexp(n, rate = 0.03)
time <- pmin(T_event, C)
status <- as.integer(T_event <= C)
y_surv <- list(time = time, status = status)
# Run VS_method
VS_method_SURV(y_surv, X, surv_model = "COX", vsel_method = "COXGLMNET", verbose = FALSE)

Package 'S3VS'

Help Index

Structured Screen-and-Select Variable Selection in Linear, Generalized Linear, and Survival Models

Description

Details

Author(s)

Bridge-Penalized AFT Regression via Iteratively Reweighted LASSO

Description

Usage

Arguments

Value

Author(s)

References

See Also

Examples

Identify Leading Sets of Covariates via Inter-Predictor Associations

Description

Usage

Arguments

Value

Author(s)

Examples

Screening Predictors As 'Leading Variables' By Evaluating Predictor-Response Associations

Description

Usage

Arguments

Value

Author(s)

See Also

Examples

Screening Predictors As 'Leading Variables' By Evaluating Predictor-Response Associations In Generalized Linear Models

Description

Usage

Arguments

Value

Author(s)

Examples

Screening Predictors As 'Leading Variables' By Evaluating Predictor-Response Associations In Linear Models

Description

Usage

Arguments

Value

Author(s)

Examples

Screening Predictors As "Leading Variables" By Evaluating Predictor-Response Associations In Survival Models

Description

Usage

Arguments

Value

Author(s)

Examples

Decide Whether to Run Another S3VS Iteration

Description

Usage

Arguments

Details

Value

Author(s)

Examples

Prediction Using S3VS-Selected Predictors

Description

Usage

Arguments

Value

Author(s)

See Also

Examples

Prediction Using S3VS-Selected Predictors in Survival Models

Description

Usage

Arguments

Value

Author(s)

See Also

Examples

Prediction Using S3VS-Selected Predictors in Linear Models

Description

Usage

Arguments

Value