Skip to contents

Normalise the decision options of an analysis-plan block

Source: R/decision_data.R
sframe_decision_options.Rd

A researcher-supplied performance matrix round-trips through JSON as a list of numeric row vectors with its dimnames dropped, so it is stored as options$matrix plus the options$alternatives and options$criteria label vectors and rebuilt here. Every length agreement between the matrix, its labels, the weights, and the criterion types is checked once, in one place, with the exact mismatch named.

Usage

sframe_decision_options(options)

Arguments

options

The options list of a decision analysis block.

Value

The same list with matrix rebuilt as a numeric matrix carrying dimnames, and weights and criteria_types coerced and checked.

PROMETHEE preference functions

A PROMETHEE block takes options$preference_function, one of "usual", "linear", or "level", and options$thresholds for the 2 that need them. The default is "usual", Brans and Vincke's type I step function, which needs no thresholds and so adds no researcher degrees of freedom.

This default differs from several other MCDM implementations, which default to the linear (V-shape) function and derive its thresholds from the range of the supplied data. Deriving thresholds that way makes the result depend on choices the researcher never declared, which is what this package exists to prevent, so surveyframe requires a threshold-bearing preference function to be asked for explicitly.

The choice changes the answer. Net flows always differ between the 2 functions, and the ranking itself changed in 226 of 400 randomly drawn 4-alternative by 3-criterion matrices. So a ranking cross-checked against an implementation that defaults to "linear" will often disagree unless preference_function = "linear" is set here and the same thresholds are supplied on both sides. The preference function actually used is always named in the block's APA sentence. Note also that "usual" produces tied ranks readily, because a step function scores every non-zero difference identically.

See also

run_analysis_plan()

Examples

sframe_decision_options(list(
  matrix = list(c(4, 210), c(3, 180)),
  alternatives = c("Alpha", "Basilica"),
  criteria = c("service", "price"),
  criteria_types = c("benefit", "cost")
))$matrix
#>          service price
#> Alpha          4   210
#> Basilica       3   180