Skip to contents

Set Coefficients for Model Objects

Source: R/mvn_helpers.R
set_coef.Rd

Replaces the coefficients in a fitted model object. This is used for simulation-based inference (MVN simulation) where we draw coefficient vectors from a multivariate normal distribution and need to compute predictions using each draw.

Usage

set_coef(model, new_coefs)

# Default S3 method
set_coef(model, new_coefs)

# S3 method for class 'orm'
set_coef(model, new_coefs)

# S3 method for class 'lrm'
set_coef(model, new_coefs)

# S3 method for class 'rms'
set_coef(model, new_coefs)

# S3 method for class 'robcov_vglm'
set_coef(model, new_coefs)

# S3 method for class 'vglm'
set_coef(model, new_coefs)

Arguments

model

A fitted model object (e.g., vglm, orm, lrm).

new_coefs

A numeric vector of coefficients to set. Must match the length and order of the original coefficients.

Value

The model object with updated coefficients.

Details

This function creates a shallow copy of the model and replaces its coefficients. The modified model can then be passed to predict() or soprob_markov() to compute predictions under the simulated coefficients.

Supported model classes:

  • orm, lrm, rms (S3): Modifies $coefficients

  • vglm (S4): Modifies @coefficients slot

  • robcov_vglm: Modifies $coefficients

Note: Prediction functions (predict.orm, predict.vglm) recompute linear predictors from the design matrix and coefficients, so we only need to replace the coefficients - not any stored linear predictors.

See also

get_vcov_robust() for obtaining robust covariance matrices

Examples

trial <- sim_markov("actt2", n_patients = 40, follow_up_time = 6, seed = 1)
markov_data <- prepare_markov_data(trial, absorbing_state = 8)
fit <- rms::orm(
  y ~ time + tx + yprev,
  data = markov_data,
  x = TRUE,
  y = TRUE,
  opt_method = "LM",
  scale = TRUE
)

new_coefs <- coef(fit)
new_coefs["tx"] <- new_coefs["tx"] - 0.1
fit_shifted <- set_coef(fit, new_coefs)
coef(fit_shifted)["tx"]
#>         tx 
#> -0.2647912