strategy_to_Ab: Transform Pattern of Predictions to Polytope

In multinomineq: Bayesian Inference for Multinomial Models with Inequality Constraints

Transform Pattern of Predictions to Polytope

Description

Transforms ordered item-type predictions to polytope definition. This allows to use Monte-Carlo methods to compute the Bayes factor if the number of item types is large (bf_binom).

Usage

strategy_to_Ab(strategy)

Arguments

strategy

a decision strategy returned by strategy_multiattribute.

Details

Note: Only works for models without guessing predictions and without equality constraints (i.e., requires separate error probabilities per item type)

Value

a list containing the matrix A and the vector b that define a polytope via A*x <= b.

Examples

# strategy:  A,B,B,A   e2<e3<e4<e1<.50
strat <- list(
  pattern = c(-1, 4, 3, -2),
  c = .5, ordered = TRUE,
  prior = c(1, 1)
)
pt <- strategy_to_Ab(strat)
pt

# compare results to encompassing BF method:
b <- list(
  pattern = 1:4, c = 1,
  ordered = FALSE, prior = c(1, 1)
)
k <- c(2, 20, 18, 0)
n <- rep(20, 4)
m1 <- strategy_postprob(k, n, list(strat, b))
log(m1[1] / m1[2])
bf_binom(k, n, pt$A, pt$b, log = TRUE)

multinomineq documentation built on May 29, 2024, 6:23 a.m.