bayes: Bayesian integration where the reliability of the advisor is...
In oxacclab/adviseR: Advice Exchange Network Modelling

Bayesian integration where the reliability of the advisor is the probability the advisor agrees given we were correct.

1	bayes(initial, advice, weight, compression = 0.05)

`initial`	vector of initial decisions
`advice`	vector of advisory estimates
`weight`	trust rating for the advisor
`compression`	whether to limit extreme values to c(x,1-x)

Uses a Bayesian integration formula where

c2 = (c1*t)/(c1*t + (1-c1)(1-t))

c2 is the final confidence (returned as a vector), and c1 the initial confidence. t is the probability of the advisor's advice given the initial decision was correct. Where the advisor agrees, this is simply the trust we have in the advisor (an advisor we trusted 100% would always be expected to give the same answer we did). Where the advisor disagrees, this is the opposite (we consider it very unlikely a highly trusted advisor disagrees with us if we are right).

oxacclab/adviseR documentation built on Oct. 7, 2021, 8:05 p.m.