simulation_conformal_scores_mix_reg: calculate some conformal scores related to simulation based...

In benjaminleroy/simulationBands: Conformal Prediction Bands with Simulations

Description

#' this wrapper assumes a model has been fit relative to the 2 mode mixture of linear regression (similar to flexmix) for the underlying mass non-conformal mass scores

Usage

simulation_conformal_scores_mix_reg(
  x,
  y,
  model_params,
  sim_num = 100,
  h = 0.4,
  grid = seq(-15, 15, by = 0.01)
)

Arguments

x	true x
y	true y
model_params	list of model parameters of with names as follows beta: a vector, length 2 intercept: a vector, length 2 sd: a vector, length 2 proportions: a vector, length 2 that represent a conditional distribution defined by y ~ p1 M1(x) + p2 M2(x) where (M1,M2)(x) ~ right_intercept + N(right_beta*x, right_sd) and (p1,p2) = right_proportions
sim_num	number of simulations to base the kernel density estimate of the cde
h	kernel density estimate bandwidth parameter
grid	discretizes grid that y should range over (needs to be equally spaced)

Value

a data.frame with 3 columns (and 1 row)

• non_conformal_mass_full_sim_based:Estimated amount of mass with CDE above y value using just the simulations.

• empirical_cde:Estimated CDE value.

• non_conformal_mass_ecde_and_sim_based:Estimated amount of mass with CDE above y value (using grid)

benjaminleroy/simulationBands documentation built on Dec. 19, 2021, 8:41 a.m.