Type: Package
Title: R Package for Pairwise Preference Elicitation
Version: 0.1.3
Date: 2022-04-24
Author: John Lepird
Maintainer: John Lepird <jlepird@alum.mit.edu>
Description: Allows users to derive multi-objective weights from pairwise comparisons, which research shows is more repeatable, transparent, and intuitive other techniques. These weights can be rank existing alternatives or to define a multi-objective utility function for optimization.
License: MIT + file LICENSE
Imports: mcmc, methods, entropy
Suggests: testthat, knitr, rmarkdown
VignetteBuilder: knitr
RoxygenNote: 7.1.2
Encoding: UTF-8
URL: https://github.com/jlepird/prefeR, https://jlepird.github.io/prefeR/
NeedsCompilation: no
Packaged: 2022-04-24 22:51:34 UTC; jack
Repository: CRAN
Date/Publication: 2022-04-24 23:00:02 UTC

A helper function to add in preferences in a user-friendly way.

Description

A helper function to add in preferences in a user-friendly way.

Usage

a %>% b

Arguments

a

The preferred row

b

The non-preferred row

See Also

Other preferences: %<%(), %=%()

Examples

1 %>% 2 # prefer row 1 to row 2

A helper function to add in preferences in a user-friendly way.

Description

A helper function to add in preferences in a user-friendly way.

Usage

a %<% b

Arguments

a

The non-preferred row

b

The preferred row

See Also

Other preferences: %=%(), %>%()

Examples

1 %<% 2 # prefer row 2 to row 1

A helper function to add in preferences in a user-friendly way.

Description

A helper function to add in preferences in a user-friendly way.

Usage

a %=% b

Arguments

a

The first alternative

b

The second alternative

See Also

Other preferences: %<%(), %>%()

Examples

1 %=% 2 # indifferent between 1 and 2

Calculates the log probability of seeing a given set of preferences

Description

Calculates the log probability of seeing a given set of preferences

Usage

.calculateLogProb(x, p)

Arguments

x

A guess for our weight vector

p

An object of the Bayes preference class

Value

A scalar log-likelihood of the guess x

Calculates the expected posterior entropy of the prefel object if x and y are compared. Ignores the odds of indifference preferences, as using them would increase runtime 50% without much gain.

Description

Calculates the expected posterior entropy of the prefel object if x and y are compared. Ignores the odds of indifference preferences, as using them would increase runtime 50% without much gain.

Usage

.estimateEntropy(p, currentGuess, x, y)

Arguments

p

An object of class BayesPrefClass.

currentGuess

The current best estimate for our weight vector.

x

Possible comparison 1

y

Possible comparison 2

Evaluates the likelihood of the observed indifference preferences

Description

Evaluates the likelihood of the observed indifference preferences

Usage

.getLogIndifProb(x, pref, p)

Arguments

x

the underlying data

pref

the stated preference

p

the preference elication object

Evaluates the likelihood of the observed strict preferences

Description

Evaluates the likelihood of the observed strict preferences

Usage

.getLogStrictProb(x, pref, p)

Arguments

x

the underlying data

pref

the stated preference

p

the preference elication object

Calculates the entropy of a matrix of samples.

Description

Calculates the entropy of a matrix of samples.

Usage

.sampleEntropy(X)

Arguments

X

a matrix where each row is a sample of variables in different columns

An object containing all data necessary for preference elicitation.

Description

An object containing all data necessary for preference elicitation.

Fields

data

A matrix or dataframe of data.

priors

A list of functions that give the prior on each variable.

sigma

A scalar value to use for the confusion factor (default 0.1).

Sigma

(Internal use only) A matrix of sigma * diag(ncol(data)).

strict

A list of lists of preferences. For each element x, x[[1]] > x[[2]].

indif

A list of lists of indifference preferences. For each element x, x[[1]] = x[[2]].

weights

A vector of weights determined by the inference algorithm.

Methods

addPref(x)

Adds a preference created using %>%, %<%, or %=%.

infer(estimate = "recommended")

Calls the “infer” function to guess weights

rank()

Calculates the utility of each row in our dataset

suggest(maxComparisons = 10)

Calls the “suggest” function to guess weights

A convenience function for generating Exponential priors.

Description

A convenience function for generating Exponential priors.

Usage

Exp(mu = 1)

Arguments

mu

The mean of the exponential distribution, i.e. 1/rate

Value

A function yielding the log-PDF at x of a exponential distribution with given statistics.

See Also

Other priors: Flat(), Normal()

Examples

Exp(1)(1) == dexp(1,1, log = TRUE)

A convenience function for generating a flat prior.

Description

A convenience function for generating a flat prior.

Usage

Flat()

Value

The zero function.

See Also

Other priors: Exp(), Normal()

Examples

Flat()(1) == 0.0

A convenience function for generating Normal priors.

Description

A convenience function for generating Normal priors.

Usage

Normal(mu = 0, sigma = 1)

Arguments

mu

The mean of the normal distribution

sigma

The standard deviation of the prior

Value

A function yielding the log-PDF at x of a normal distribution with given statistics.

See Also

Other priors: Exp(), Flat()

Examples

Normal(0, 1)(1) == dnorm(1, log = TRUE)

A function that estimates the user's underlying utility function.

Description

A function that estimates the user's underlying utility function.

Usage

infer(p, estimate = "recommended", nbatch = 1000)

Arguments

p

A BayesPrefClass instance.

estimate

The type of posterior point-estimate returned. Valid options are "recommended" (default), "MAP", and "mean".

nbatch

If using Monte Carlo estimates, the number of samples. Defaults to 1000.

Value

A vector of parameters that best fits the observed preferences.

Examples

 p <- prefEl(data = data.frame(c(1,0,1), c(0,1,1), c(1,1,1)),
             priors = c(Normal(0, 1), Exp(0.5), Flat()))
 p$addPref(1 %>% 2)
 infer(p, estimate = "MAP")

A shortcut to create objects of the class BayesPrefClass.

Description

A shortcut to create objects of the class BayesPrefClass.

Usage

prefEl(data = NA, priors = list(), ...)

Arguments

data

A matrix or dataframe of data. Each column should be a variable, each row an observation.

priors

A list of functions that give the prior on each variable. E.g. see help(Flat)

...

Other parameters to pass to the class constructor. Not recommended.

Examples

p <- prefEl(data = data.frame(x = c(1,0,1), y = c(0, 1, 1)),
            priors = c(Normal(0,1), Flat()))

Suggests a good comparison for the user to make next.

Description

Suggests a good comparison for the user to make next.

Usage

suggest(p, maxComparisons = 10)

Arguments

p

An object of class BayesPrefClass.

maxComparisons

The maximum number of possible comparisons to check. Default: 10.

Value

A two-element vector of recommended comparisons.