startthat.work home

PJ-01 Trend extrapolation and time-series forecasting (exponential smoothing, autoregressive integrated moving average models, trend impact analysis)

Extend the measured history of one indicator with a stated functional form to give a central projection with prediction intervals, as a draft result.

Project forwardEvidence: UnconfirmedPrimary technique, runs in step 5.3Strongest statement: predict, without claiming a cause

What it does

Extends the measured historical path of an indicator using a stated functional form. Artefact: a central projection with prediction intervals; read interval width and compare with a naive no-change forecast.

When to use it

Use when you need a forward projection of a stable quantity that has a long, consistently defined history and a stated horizon.

When it does not fit

The series is short or its definition changed, the projection does not beat a naive benchmark in back-testing, no interval is stated, or the driver of the trend is reaching a limit.

Result format

The result is a central projection of the indicator with prediction intervals, based on a stated functional form. It gives the interval width and a comparison with a naive no-change forecast.

Requirements on the material

PropertyRequirement typeRequirement
StructureApplies in some casesrequires=structured; requires=semi-structured
Content typeAlways appliesrequires=time series
Unit of analysisAlways appliesrequires=variable
Variable typesRaises a warningprefers=ratio; prefers=interval; prefers=counts or fractions
Time structureApplies in some casesrequires=aggregate time series; requires=repeated cross-section
Level of detail availableAlways appliesrequires=aggregate tables
Appraises source credibilityApplies in some casesno

Not restricted on: 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 materialA series with a constant definition and enough periods to cover at least two full seasonal cycles where seasonality exists, plus a stated horizon.
Additional sources needednone
Not for (rules)none
Not for (text)The series is short or its definition changed, no interval is stated, or the driver of the trend is reaching a limit.

Choosing and ordering

Kind of techniquePrimary technique, runs in step 5.3
Phase taganalyse
Other phase tagsnone
Kind of findingtrend; risk
Question answeredwhat next; how much
Strongest statement it supportspredict, without claiming a cause
Needs as input
Makes as output
Must exist firstNone
Loop targetreturn to analyse

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
decision-informinginput validationnoneno
decision-informingheld-out validationnoneno
decision-informingsensitivity analysisnoneno
high-stakesinput validationnoneno
high-stakesheld-out validationnoneno
high-stakessensitivity analysisnoneno
high-stakesleakage checknoneno
high-stakessecond technique with different assumptionsnoneconditional
high-stakesassumption and limitation statementnoneno

Failure modes and limits

Model suitabilitymodel plus code
Typical model failureExtends the trend by eye or in text arithmetic and omits the prediction interval and the comparison with a naive no-change forecast.
Error sourcesinput data; statistical assumptions
Certainty limitsCertainty drops when back-testing does not beat a naive benchmark (indirectness), the series is short or changed definition (risk of bias), or the interval is wide (imprecision).

Evidence and tag basis

Unconfirmed. Forecasting competition results and the Principles of Forecasting were recalled, and the page failed to load. The note author expects the strongest accuracy base here, limited to stable quantities, but this was not verified.

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