Help for package tramvs

Type:

Package

Title:

Optimal Subset Selection for Transformation Models

Version:

0.0-7

Description:

Greedy optimal subset selection for transformation models (Hothorn et al., 2018, <doi:10.1111/sjos.12291> ) based on the abess algorithm (Zhu et al., 2020, <doi:10.1073/pnas.2014241117> ). Applicable to models from packages 'tram' and 'cotram'. Application to shift-scale transformation models are described in Siegfried et al. (2024, <doi:10.1080/00031305.2023.2203177>).

Depends:

R (≥ 4.2.0), tram (≥ 0.6-1)

Imports:

stats, variables, methods, cotram, future, future.apply, mvtnorm

Suggests:

abess, tramnet, colorspace, knitr, mlt, TH.data, survival, ordinal, testthat (≥ 3.0.0)

License:

GPL-3

Encoding:

UTF-8

RoxygenNote:

7.3.1

URL:

http://ctm.R-forge.R-project.org

VignetteBuilder:

knitr

Config/testthat/edition:

NeedsCompilation:

Packaged:

2025-01-31 09:40:30 UTC; lkook

Author:

Lucas Kook

[aut, cre], Sandra Siegfried [ctb], Torsten Hothorn [ctb]

Maintainer:

Lucas Kook <lucasheinrich.kook@gmail.com>

Repository:

CRAN

Date/Publication:

2025-01-31 10:10:02 UTC

AIC "tramvs"

Description

AIC "tramvs"

Usage

## S3 method for class 'tramvs'
AIC(object, ...)

Arguments

object

object of class "tramvs"

...

additional arguments to AIC()

Value

Numeric vector containing AIC of best model

Optimal subset selection in a BoxCox-type transformation model

Description

Optimal subset selection in a BoxCox-type transformation model

Usage

BoxCoxVS(
  formula,
  data,
  supp_max = NULL,
  k_max = NULL,
  thresh = NULL,
  init = TRUE,
  m_max = 10,
  parallel = FALSE,
  future_args = list(strategy = "multisession", workers = supp_max),
  ...
)

Arguments

formula

object of class "formula".

data

data frame containing the variables in the model.

supp_max

maximum support which to call abess_tram with.

k_max

maximum support size to consider during the splicing algorithm. Defaults to supp.

thresh

threshold when to stop splicing. Defaults to 0.01 * supp * p * log(log(n)) / n$, where p denotes the number of predictors and n the sample size.

init

initialize active set. Defaults to TRUE and initializes the active set with those covariates that are most correlated with score residuals of an unconditional modFUN(update(formula, . ~ 1)).

m_max

maximum number of iterating the splicing algorithm.

parallel

toggle for parallel computing via future_lapply

future_args

arguments passed to plan; defaults to a "multisession" with supp_max workers

...

Additional arguments supplied to BoxCox

Value

See tramvs

Optimal subset selection in a Colr-type transformation model

Description

Optimal subset selection in a Colr-type transformation model

Usage

ColrVS(
  formula,
  data,
  supp_max = NULL,
  k_max = NULL,
  thresh = NULL,
  init = TRUE,
  m_max = 10,
  parallel = FALSE,
  future_args = list(strategy = "multisession", workers = supp_max),
  ...
)

Arguments

formula

object of class "formula".

data

data frame containing the variables in the model.

supp_max

maximum support which to call abess_tram with.

k_max

maximum support size to consider during the splicing algorithm. Defaults to supp.

thresh

threshold when to stop splicing. Defaults to 0.01 * supp * p * log(log(n)) / n$, where p denotes the number of predictors and n the sample size.

init

initialize active set. Defaults to TRUE and initializes the active set with those covariates that are most correlated with score residuals of an unconditional modFUN(update(formula, . ~ 1)).

m_max

maximum number of iterating the splicing algorithm.

parallel

toggle for parallel computing via future_lapply

future_args

arguments passed to plan; defaults to a "multisession" with supp_max workers

...

Additional arguments supplied to Colr

Value

See tramvs

Optimal subset selection in a Coxph-type transformation model

Description

Optimal subset selection in a Coxph-type transformation model

Usage

CoxphVS(
  formula,
  data,
  supp_max = NULL,
  k_max = NULL,
  thresh = NULL,
  init = TRUE,
  m_max = 10,
  parallel = FALSE,
  future_args = list(strategy = "multisession", workers = supp_max),
  ...
)

Arguments

formula

object of class "formula".

data

data frame containing the variables in the model.

supp_max

maximum support which to call abess_tram with.

k_max

maximum support size to consider during the splicing algorithm. Defaults to supp.

thresh

threshold when to stop splicing. Defaults to 0.01 * supp * p * log(log(n)) / n$, where p denotes the number of predictors and n the sample size.

init

initialize active set. Defaults to TRUE and initializes the active set with those covariates that are most correlated with score residuals of an unconditional modFUN(update(formula, . ~ 1)).

m_max

maximum number of iterating the splicing algorithm.

parallel

toggle for parallel computing via future_lapply

future_args

arguments passed to plan; defaults to a "multisession" with supp_max workers

...

Additional arguments supplied to Coxph

Value

See tramvs

Optimal subset selection in a Lehmann-type transformation model

Description

Optimal subset selection in a Lehmann-type transformation model

Usage

LehmannVS(
  formula,
  data,
  supp_max = NULL,
  k_max = NULL,
  thresh = NULL,
  init = TRUE,
  m_max = 10,
  parallel = FALSE,
  future_args = list(strategy = "multisession", workers = supp_max),
  ...
)

Arguments

formula

object of class "formula".

data

data frame containing the variables in the model.

supp_max

maximum support which to call abess_tram with.

k_max

maximum support size to consider during the splicing algorithm. Defaults to supp.

thresh

threshold when to stop splicing. Defaults to 0.01 * supp * p * log(log(n)) / n$, where p denotes the number of predictors and n the sample size.

init

initialize active set. Defaults to TRUE and initializes the active set with those covariates that are most correlated with score residuals of an unconditional modFUN(update(formula, . ~ 1)).

m_max

maximum number of iterating the splicing algorithm.

parallel

toggle for parallel computing via future_lapply

future_args

arguments passed to plan; defaults to a "multisession" with supp_max workers

...

Additional arguments supplied to Lehmann

Value

See tramvs

Optimal subset selection in an Lm-type transformation model

Description

Optimal subset selection in an Lm-type transformation model

Usage

LmVS(
  formula,
  data,
  supp_max = NULL,
  k_max = NULL,
  thresh = NULL,
  init = TRUE,
  m_max = 10,
  parallel = FALSE,
  future_args = list(strategy = "multisession", workers = supp_max),
  ...
)

Arguments

formula

object of class "formula".

data

data frame containing the variables in the model.

supp_max

maximum support which to call abess_tram with.

k_max

maximum support size to consider during the splicing algorithm. Defaults to supp.

thresh

threshold when to stop splicing. Defaults to 0.01 * supp * p * log(log(n)) / n$, where p denotes the number of predictors and n the sample size.

init

initialize active set. Defaults to TRUE and initializes the active set with those covariates that are most correlated with score residuals of an unconditional modFUN(update(formula, . ~ 1)).

m_max

maximum number of iterating the splicing algorithm.

parallel

toggle for parallel computing via future_lapply

future_args

arguments passed to plan; defaults to a "multisession" with supp_max workers

...

Additional arguments supplied to Lm

Value

See tramvs

Optimal subset selection in a Polr-type transformation model

Description

Optimal subset selection in a Polr-type transformation model

Usage

PolrVS(
  formula,
  data,
  supp_max = NULL,
  k_max = NULL,
  thresh = NULL,
  init = TRUE,
  m_max = 10,
  parallel = FALSE,
  future_args = list(strategy = "multisession", workers = supp_max),
  ...
)

Arguments

formula

object of class "formula".

data

data frame containing the variables in the model.

supp_max

maximum support which to call abess_tram with.

k_max

maximum support size to consider during the splicing algorithm. Defaults to supp.

thresh

threshold when to stop splicing. Defaults to 0.01 * supp * p * log(log(n)) / n$, where p denotes the number of predictors and n the sample size.

init

initialize active set. Defaults to TRUE and initializes the active set with those covariates that are most correlated with score residuals of an unconditional modFUN(update(formula, . ~ 1)).

m_max

maximum number of iterating the splicing algorithm.

parallel

toggle for parallel computing via future_lapply

future_args

arguments passed to plan; defaults to a "multisession" with supp_max workers

...

Additional arguments supplied to Polr

Value

See tramvs

SIC generic

Description

SIC generic

Usage

SIC(object, ...)

Arguments

object

Model to compute SIC from

...

for methods compatibility only

Value

Numeric vector (best_only = TRUE) or data.frame with SIC values

SIC "tramvs"

Description

SIC "tramvs"

Usage

## S3 method for class 'tramvs'
SIC(object, best_only = FALSE, ...)

Arguments

object

object of class "tramvs"

best_only

Wether to return the coefficients of the best model only (default: FALSE)

...

for methods compatibility only

Value

Numeric vector (best_only = TRUE) or data.frame with SIC values

Optimal subset selection in a Survreg model

Description

Optimal subset selection in a Survreg model

Usage

SurvregVS(
  formula,
  data,
  supp_max = NULL,
  k_max = NULL,
  thresh = NULL,
  init = TRUE,
  m_max = 10,
  parallel = FALSE,
  future_args = list(strategy = "multisession", workers = supp_max),
  ...
)

Arguments

formula

object of class "formula".

data

data frame containing the variables in the model.

supp_max

maximum support which to call abess_tram with.

k_max

maximum support size to consider during the splicing algorithm. Defaults to supp.

thresh

threshold when to stop splicing. Defaults to 0.01 * supp * p * log(log(n)) / n$, where p denotes the number of predictors and n the sample size.

init

initialize active set. Defaults to TRUE and initializes the active set with those covariates that are most correlated with score residuals of an unconditional modFUN(update(formula, . ~ 1)).

m_max

maximum number of iterating the splicing algorithm.

parallel

toggle for parallel computing via future_lapply

future_args

arguments passed to plan; defaults to a "multisession" with supp_max workers

...

Additional arguments supplied to Survreg

Value

See tramvs

Optimal subset selection for multivariate transformation models

Description

Optimal subset selection for multivariate transformation models

Usage

abess_mmlt(
  mltargs,
  supp,
  k_max = supp,
  thresh = NULL,
  init = TRUE,
  m_max = 10,
  m0 = NULL,
  ...
)

Arguments

mltargs

Arguments passed to mmlt

supp

support size of the coefficient vector

k_max

maximum support size to consider during the splicing algorithm. Defaults to supp.

thresh

threshold when to stop splicing. Defaults to 0.01 * supp * p * log(log(n)) / n$, where p denotes the number of predictors and n the sample size.

init

initialize active set. Defaults to TRUE and initializes the active set with those covariates that are most correlated with score residuals of an unconditional modFUN(update(formula, . ~ 1)).

m_max

maximum number of iterating the splicing algorithm.

m0

Transformation model for initialization

...

Currently ignored

Value

List containing the fitted model via mmlt, active set A and inactive set I.

Optimal subset selection for transformation models

Description

Optimal subset selection for transformation models

Usage

abess_tram(
  formula,
  data,
  modFUN,
  supp,
  mandatory = NULL,
  k_max = supp,
  thresh = NULL,
  init = TRUE,
  m_max = 10,
  m0 = NULL,
  ...
)

Arguments

formula

object of class "formula".

data

data frame containing the variables in the model.

modFUN

function for fitting a transformation model, e.g., BoxCox().

supp

support size of the coefficient vector

mandatory

formula of mandatory covariates, which will always be included and estimated in the model. Note that this also changes the intialization of the active set. The active set is then computed with regards to the model residuals of modFUN(mandatory, ...) instead of the unconditional model.

k_max

maximum support size to consider during the splicing algorithm. Defaults to supp.

thresh

threshold when to stop splicing. Defaults to 0.01 * supp * p * log(log(n)) / n$, where p denotes the number of predictors and n the sample size.

init

initialize active set. Defaults to TRUE and initializes the active set with those covariates that are most correlated with score residuals of an unconditional modFUN(update(formula, . ~ 1)).

m_max

maximum number of iterating the splicing algorithm.

m0

Transformation model for initialization

...

additional arguments supplied to modFUN.

Value

List containing the fitted model via modFUN, active set A and inactive set I.

Examples

set.seed(24101968)
library(tramvs)

N <- 1e2
P <- 5
nz <- 3
beta <- rep(c(1, 0), c(nz, P - nz))
X <- matrix(rnorm(N * P), nrow = N, ncol = P)
Y <- 1 + X %*% beta + rnorm(N)

dat <- data.frame(y = Y, x = X)

abess_tram(y ~ ., dat, modFUN = Lm, supp = 3)

Coef "abess_tram"

Description

Coef "abess_tram"

Usage

## S3 method for class 'abess_tram'
coef(object, ...)

Arguments

object

object of class "tramvs"

...

additional arguments to coef()

Value

Named numeric vector containing coefficient estimates see coef.tram

Coef "mmltvs"

Description

Coef "mmltvs"

Usage

## S3 method for class 'mmltvs'
coef(object, best_only = FALSE, ...)

Arguments

object

Object of class "tramvs"

best_only

Wether to return the coefficients of the best model only (default: FALSE)

...

additional arguments to coef()

Value

Vector (best_only = TRUE) or matrix (best_only = FALSE) of coefficients

Coef "tramvs"

Description

Coef "tramvs"

Usage

## S3 method for class 'tramvs'
coef(object, best_only = FALSE, ...)

Arguments

object

Object of class "tramvs"

best_only

Wether to return the coefficients of the best model only (default: FALSE)

...

additional arguments to coef()

Value

Vector (best_only = TRUE) or matrix (best_only = FALSE) of coefficients

Compute correlation for initializing the active set

Description

Compute correlation for initializing the active set

Usage

cor_init(m0, mb)

Arguments

m0

modFUN(formula, data)

mb

modFUN(mandatory, data)

Value

Vector of correlations for initializing the active set, depends on type of model (see e.g. cor_init.default)

Default method for computing correlation

Description

Default method for computing correlation

Usage

## Default S3 method:
cor_init(m0, mb)

Arguments

m0

modFUN(formula, data)

mb

modFUN(mandatory, data)

Value

Vector of correlation for initializing the active set

Method for computing correlations in mmlts

Description

Method for computing correlations in mmlts

Usage

## S3 method for class 'mmlt'
cor_init(m0, mb)

Arguments

m0

modFUN(formula, data)

mb

modFUN(mandatory, data)

Value

Vector of correlation for initializing the active set

Shit-scale tram method for computing correlation

Description

Shit-scale tram method for computing correlation

Usage

## S3 method for class 'stram'
cor_init(m0, mb)

Arguments

m0

modFUN(formula, data)

mb

modFUN(mandatory, data)

Value

Vector of correlations for initializing the active set, includes both shift and scale residuals

Optimal subset selection in a cotram model

Description

Optimal subset selection in a cotram model

Usage

cotramVS(
  formula,
  data,
  supp_max = NULL,
  k_max = NULL,
  thresh = NULL,
  init = TRUE,
  m_max = 10,
  parallel = FALSE,
  future_args = list(strategy = "multisession", workers = supp_max),
  ...
)

Arguments

formula

object of class "formula".

data

data frame containing the variables in the model.

supp_max

maximum support which to call abess_tram with.

k_max

maximum support size to consider during the splicing algorithm. Defaults to supp.

thresh

threshold when to stop splicing. Defaults to 0.01 * supp * p * log(log(n)) / n$, where p denotes the number of predictors and n the sample size.

init

initialize active set. Defaults to TRUE and initializes the active set with those covariates that are most correlated with score residuals of an unconditional modFUN(update(formula, . ~ 1)).

m_max

maximum number of iterating the splicing algorithm.

parallel

toggle for parallel computing via future_lapply

future_args

arguments passed to plan; defaults to a "multisession" with supp_max workers

...

Additional arguments supplied to cotram

Value

See tramvs

logLik "tramvs"

Description

logLik "tramvs"

Usage

## S3 method for class 'tramvs'
logLik(object, ...)

Arguments

object

object of class "tramvs"

...

additional arguments to logLik()

Value

Numeric vector containing log-likelihood of best model, see logLik.tram

Select optimal subset based on high dimensional BIC in mmlts

Description

Select optimal subset based on high dimensional BIC in mmlts

Usage

mmltVS(
  mltargs,
  supp_max = NULL,
  k_max = NULL,
  thresh = NULL,
  init = TRUE,
  m_max = 10,
  verbose = TRUE,
  parallel = FALSE,
  m0 = NULL,
  future_args = list(strategy = "multisession", workers = supp_max),
  ...
)

Arguments

mltargs

Arguments passed to mmlt

supp_max

maximum support which to call abess_tram with.

k_max

maximum support size to consider during the splicing algorithm. Defaults to supp.

thresh

threshold when to stop splicing. Defaults to 0.01 * supp * p * log(log(n)) / n$, where p denotes the number of predictors and n the sample size.

init

initialize active set. Defaults to TRUE and initializes the active set with those covariates that are most correlated with score residuals of an unconditional modFUN(update(formula, . ~ 1)).

m_max

maximum number of iterating the splicing algorithm.

verbose

show progress bar (default: TRUE)

parallel

toggle for parallel computing via future_lapply

m0

Transformation model for initialization

future_args

arguments passed to plan; defaults to a "multisession" with supp_max workers

...

Arguments passed on to abess_mmlt

supp: support size of the coefficient vector

Details

L0-penalized (i.e., best subset selection) multivariate transformation models using the abess algorithm.

Value

object of class "mltvs", containing the regularization path (information criterion SIC and coefficients coefs), the best fit (best_fit) and all other models (all_fits)

Plot "tramvs" object

Description

Plot "tramvs" object

Usage

## S3 method for class 'tramvs'
plot(x, which = c("tune", "path"), ...)

Arguments

x

object of class "tramvs"

which

plotting either the regularization path ("path") or the information criterion against the support size ("tune", default)

...

additional arguments to plot()

Value

Returns invisible(NULL)

Predict "tramvs"

Description

Predict "tramvs"

Usage

## S3 method for class 'tramvs'
predict(object, ...)

Arguments

object

object of class "tramvs"

...

additional arguments to predict.tram()

Value

See predict.tram

Print "tramvs"

Description

Print "tramvs"

Usage

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

Arguments

x

object of class "tramvs"

...

ignored

Value

"tramvs" object is returned invisibly

Residuals "tramvs"

Description

Residuals "tramvs"

Usage

## S3 method for class 'tramvs'
residuals(object, ...)

Arguments

object

object of class "tramvs"

...

additional arguments to residuals()

Value

Numeric vector containing residuals of best model, see residuals.tram

Simulate "tramvs"

Description

Simulate "tramvs"

Usage

## S3 method for class 'tramvs'
simulate(object, nsim = 1, seed = NULL, ...)

Arguments

object

object of class "tramvs"

nsim

number of simulations

seed

random seed for simulation

...

additional arguments to simulate()

Value

See simulate.mlt

Summary "tramvs"

Description

Summary "tramvs"

Usage

## S3 method for class 'tramvs'
summary(object, ...)

Arguments

object

object of class "tramvs"

...

ignored

Value

"tramvs" object is returned invisibly

Support "tramvs"

Description

Support "tramvs"

Usage

## S3 method for class 'tramvs'
support(object, ...)

Arguments

object

object of class "tramvs"

...

ignored

Value

Character vector containing active set of best fit

Select optimal subset based on high dimensional BIC

Description

Select optimal subset based on high dimensional BIC

Usage

tramvs(
  formula,
  data,
  modFUN,
  mandatory = NULL,
  supp_max = NULL,
  k_max = NULL,
  thresh = NULL,
  init = TRUE,
  m_max = 10,
  m0 = NULL,
  verbose = TRUE,
  parallel = FALSE,
  future_args = list(strategy = "multisession", workers = supp_max),
  ...
)

Arguments

formula

object of class "formula".

data

data frame containing the variables in the model.

modFUN

function for fitting a transformation model, e.g., BoxCox().

mandatory

supp_max

maximum support which to call abess_tram with.

k_max

maximum support size to consider during the splicing algorithm. Defaults to supp.

thresh

threshold when to stop splicing. Defaults to 0.01 * supp * p * log(log(n)) / n$, where p denotes the number of predictors and n the sample size.

init

initialize active set. Defaults to TRUE and initializes the active set with those covariates that are most correlated with score residuals of an unconditional modFUN(update(formula, . ~ 1)).

m_max

maximum number of iterating the splicing algorithm.

m0

Transformation model for initialization

verbose

show progress bar (default: TRUE)

parallel

toggle for parallel computing via future_lapply

future_args

arguments passed to plan; defaults to a "multisession" with supp_max workers

...

Arguments passed on to abess_tram

supp: support size of the coefficient vector

Details

L0-penalized (i.e., best subset selection) transformation models using the abess algorithm.

Value

object of class "tramvs", containing the regularization path (information criterion SIC and coefficients coefs), the best fit (best_fit) and all other models (all_fits)

Examples

set.seed(24101968)
library("tramvs")

N <- 1e2
P <- 5
nz <- 3
beta <- rep(c(1, 0), c(nz, P - nz))
X <- matrix(rnorm(N * P), nrow = N, ncol = P)
Y <- 1 + X %*% beta + rnorm(N)

dat <- data.frame(y = Y, x = X)
res <- tramvs(y ~ ., data = dat, modFUN = Lm)
plot(res, type = "b")
plot(res, which = "path")