startthat.work home

AO-13 Decision tree and expected value

Lays out decisions, chance events and outcomes and computes expected value or utility, producing a decision tree that shows the alternative with the best expected value.

Assess actors, options and risksEvidence: DocumentedPrimary technique, runs in step 5.3Strongest statement: predict, without claiming a cause

What it does

Lays out decisions, chance events and outcomes and computes expected value or utility. Artefact: a tree; read the alternative with the best expected value.

When to use it

Use when a user faces a sequence of choices and uncertain events and asks which choice has the best expected outcome.

When it does not fit

Probabilities are guesses; small data changes can change the best tree.

Result format

The result is a tree that lays out the decisions, chance events and outcomes, with the probability and value stated at each chance event and outcome. It gives the expected value of each alternative, names the alternative with the best expected value and lists the assumptions behind the figures.

Requirements on the material

PropertyRequirement typeRequirement
Content typeAlways appliesrequires=tabular
Unit of analysisAlways appliesrequires=topic
Variable typesRaises a warningprefers=counts or fractions; prefers=ratio
Level of detail availableAlways appliesrequires=aggregate tables
Appraises source credibilityApplies in some casesno

Not restricted on: Structure, Time structure, How the sample was chosen, Number of sources, Whether sources can be joined, Amount of material, Known data problems, Producer interest, Source dependence, Origin, Document genre, Language, Personal information, Security classification, Licence.

Minimum materialAt least two alternatives with outcome probabilities and costs or utilities; expert descriptions can help when data are scarce.
Additional sources needednone
Not for (rules)none
Not for (text)Wrong choice when probabilities are guesses or when small data changes would change the best branch. Large trees are hard to draw and influence diagrams are more compact.

Choosing and ordering

Kind of techniquePrimary technique, runs in step 5.3
Phase taganalyse
Other phase tagsconclude
Kind of findingoption ranking; risk
Question answeredwhich option; what next
Strongest statement it supportspredict, without claiming a cause
Needs as input
Makes as output
Must exist firstNone
Loop targetreturn to gather

Library checks by stakes

Each stakes level includes the checks of the level below it. Checks that need a person apply at high stakes only, so the process leaves them out at lower stakes (decision 38). Repeatability: deterministic.

StakesCheckPerson involvedCounts as a different method
exploratoryinput validationnoneno
exploratorycode execution for arithmeticnoneno
decision-informinginput validationnoneno
decision-informingcode execution for arithmeticnoneno
decision-informingassumption and limitation statementnoneno
decision-informingsensitivity analysisnoneno
high-stakesinput validationnoneno
high-stakescode execution for arithmeticnoneno
high-stakesassumption and limitation statementnoneno
high-stakessensitivity analysisnoneno
high-stakessecond technique with different assumptionsnoneconditional
high-stakesexpert reviewneededno

Failure modes and limits

Model suitabilityhuman judgement required
Typical model failureSupplies guessed probabilities as if measured, and computes expected values in text with arithmetic errors.
Error sourcesinput data; statistical assumptions; expert opinion
Certainty limitsCertainty drops when probabilities and utilities are elicited guesses (imprecision), when the best branch changes under small input changes (inconsistency), and when the tree omits relevant events.

Evidence and tag basis

Documented. Large trees are hard to draw and influence diagrams are more compact; the page flags an error in its own example. (Wikipedia, Decision tree)

Tag basis: input profile inference; selection inference; sequencing inference; robustness inference. All tag values are inference unless a source is named.