criterion: Compute information criteria for model comparison

In mcp: Regression with Multiple Change Points

Compute information criteria for model comparison

Description

Takes an mcpfit as input and computes information criteria using loo or WAIC. Compare models using loo_compare and loo_model_weights. more in loo.

Usage

criterion(fit, criterion = "loo", ...)

## S3 method for class 'mcpfit'
loo(x, ...)

## S3 method for class 'mcpfit'
waic(x, ...)

Arguments

fit	An mcpfit object.
criterion	One of "loo" (calls loo) or "waic" (calls waic).
...	Currently ignored
x	An mcpfit object.

Value

a loo or psis_loo object.

Functions

• loo(mcpfit): Computes loo on mcpfit objects

• waic(mcpfit): Computes WAIC on mcpfit objects

Author(s)

Jonas Kristoffer Lindeløv [e-mail]

See Also

criterion

criterion

Examples

# Define two models and sample them
# options(mc.cores = 3)  # Speed up sampling
ex = mcp_example("intercepts")  # Get some simulated data.
model1 = list(y ~ 1 + x, ~ 1)
model2 = list(y ~ 1 + x)  # Without a change point
fit1 = mcp(model1, ex$data)
fit2 = mcp(model2, ex$data)

# Compute LOO for each and compare (works for waic(fit) too)
fit1$loo = loo(fit1)
fit2$loo = loo(fit2)
loo::loo_compare(fit1$loo, fit2$loo)

mcp documentation built on April 1, 2023, 12:03 a.m.