LPDS: Log Predictive Density Score

In shrinkGPR: Scalable Gaussian Process Regression with Hierarchical Shrinkage Priors

Log Predictive Density Score

Description

LPDS calculates the log predictive density score for a fitted shrinkGPR model using a test dataset.

Usage

LPDS(mod, data_test, nsamp = 100)

Arguments

mod	A shrinkGPR object representing the fitted Gaussian process regression model.
data_test	Data frame with one row containing the covariates for the test set. Variables in data_test must match those used in model fitting.
nsamp	Positive integer specifying the number of posterior samples to use for the evaluation. Default is 100.

Details

The log predictive density score is a measure of model fit that evaluates how well the model predicts unseen data. It is computed as the log of the marginal predictive density of the observed responses.

Value

A numeric value representing the log predictive density score for the test dataset.

Examples

if (torch::torch_is_installed()) {
  # Simulate data
  set.seed(123)
  torch::torch_manual_seed(123)
  n <- 100
  x <- matrix(runif(n * 2), n, 2)
  y <- sin(2 * pi * x[, 1]) + rnorm(n, sd = 0.1)
  data <- data.frame(y = y, x1 = x[, 1], x2 = x[, 2])

  # Fit GPR model
  res <- shrinkGPR(y ~ x1 + x2, data = data)

  # Calculate true y value and calculate LPDS at specific point
  x1_new <- 0.8
  x2_new <- 0.5
  y_true <- sin(2 * pi * x1_new)
  data_test <- data.frame(y = y_true, x1 = x1_new, x2 = x2_new)
  LPDS(res, data_test)
  }

shrinkGPR documentation built on April 4, 2025, 3:07 a.m.