Type: Package
Title: 'Burgle': Stealing the Necessary Parts of Model Objects
Version: 0.1.2
Maintainer: Paul R. Gunsalus <gunsalp@ccf.org>
Description: Provides a way to reduce model objects to necessary parts, making them easier to work with, store, share and simulate multiple values for new responses while allowing for parameter uncertainty.
License: MIT + file LICENSE
Encoding: UTF-8
RoxygenNote: 7.2.3
Imports: stats, MASS, survival, riskRegression
Suggests: flexsurv, nnet
Depends: R (≥ 4.0.0)
NeedsCompilation: no
Packaged: 2024-10-01 01:11:33 UTC; gunsalp
Author: Paul R. Gunsalus ORCID iD [aut, cre], Jarrod E. Dalton ORCID iD [aut], Adam T. Perzynski ORCID iD [aut]
Repository: CRAN
Date/Publication: 2024-10-01 08:40:07 UTC

Burgle

Description

Burgling what is necessary from different objects

Usage

burgle(object, ...)

## S3 method for class 'lm'
burgle(object, ...)

## S3 method for class 'glm'
burgle(object, ...)

## S3 method for class 'CauseSpecificCox'
burgle(object, ...)

## S3 method for class 'cph'
burgle(object, ...)

## S3 method for class 'flexsurvreg'
burgle(object, ...)

## S3 method for class 'multinom'
burgle(object, ...)

## S3 method for class 'coxph'
burgle(object, ...)

Arguments

object

the model object to burgle

...

must be left empty for now

Value

a burgle_ object

Examples

fit <- lm(Sepal.Length ~ Sepal.Width + Petal.Length, data = iris)
bfit <- burgle(fit)
object.size(fit)
object.size(bfit)

Predict for burgle methods

Description

Predict for burgle methods

Usage

## S3 method for class 'burgle_CauseSpecificCox'
predict(
  object,
  newdata = NULL,
  type = "lp",
  cause = 1,
  original = TRUE,
  draws = 1,
  sims = 1,
  times = NULL,
  ...
)

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

## S3 method for class 'burgle_flexsurvreg'
predict(
  object,
  newdata = NA,
  original = TRUE,
  draws = 1,
  sims = 1,
  type = "lp",
  times = NULL,
  ...
)

## S3 method for class 'burgle_multinom'
predict(
  object,
  newdata = NA,
  original = TRUE,
  draws = 1,
  sims = 1,
  type = "lp",
  floor = FALSE,
  seed = NULL,
  ...
)

## S3 method for class 'burgle_coxph'
predict(
  object,
  newdata = NA,
  original = TRUE,
  draws = 1,
  sims = 1,
  type = "lp",
  times = NULL,
  ...
)

## S3 method for class 'burgle_lm'
predict(
  object,
  newdata,
  original = TRUE,
  draws = 1,
  sims = 1,
  type = "lp",
  se = FALSE,
  limits = NULL,
  ...
)

## S3 method for class 'burgle_glm'
predict(
  object,
  newdata,
  original = TRUE,
  draws = 1,
  sims = 1,
  type = "lp",
  se = FALSE,
  ...
)

Arguments

object

the results of burgle_* object

newdata

new data of class data.frame

type

either 'lp', 'response', 'link' for glm or 'risk' if time dependent

cause

which cause do you want to predict

original

whether or not to predict using the original model

draws

how many different models to simulate

sims

how many simulated response to draw

times

if type = "risk" time for which to predict risk, if times and sims is multiple the return will be lists within lists

...

for future methods

floor

will set the minimum odds to 0, if negative odds exists

seed

a seed to specificy for simulating responses (multinomial only)

se

whether or not to include the standard error in the simulations

limits

limits (minimum and maximum) for simulated response values.

Value

either a matrix or list of new model predictions