prin_vars: Determine Principle Outputs

In Tandethsquire/hmer: History Matching and Emulation Package

Determine Principle Outputs

Description

Given model outputs, determine those which are most informative for emulation.

Usage

prin_vars(data, max_vars = length(data), var_cut = 0.95)

Arguments

data	The model outputs
max_vars	The maximum number of outputs allowed in the output
var_cut	The desired proportion of variance explained

Details

For models with large numbers of outputs, it may not be useful or meaningful to emulate every output at every wave; particularly at early waves, the main effects can often be explained by a small number of model outputs. This function determines those outputs which contribute most highly to the model variation across the space.

Two cut-off points for informative outputs are available: max_vars allows one to limit the number of outputs emulated, while var_cut will continue to select outputs until a given proportion of variation has been explained. By default, no maximum number of variables is imposed and the desired variation explained is 95

The output is a list of two elements: the first, a set of variable names ordered by variance explained (the first being the most informative); the second a record of the cumulative variance explained upon inclusion of each of the outputs.

Value

A list list(ordered_variables, cumulative_variance)

Examples

# Simple example using SIR data
sir_data <- rbind.data.frame(SIRSample$training, SIRSample$validation)
# Selects nS, nR as informative
prin_vars(sir_data)
# Pick only the first: max_vars overrides var_cut
prin_vars(sir_data, max_vars = 1, var_cut = 1)

Tandethsquire/hmer documentation built on April 14, 2025, 12:48 p.m.