SIGNALAF

Transportability

0Concept

Whether metrics and findings from one cohort or context transfer to another. Can results from one population generalize?

Definition

Whether metrics and findings from one cohort or context transfer to another. Can results from one population generalize?

Inputs

Metric distributions from two or more cohorts/contexts.

Derived variables

Distribution overlap, transfer error rate.

Claim

If metrics are transportable, findings from one cohort should generalize to another (within reasonable bounds).

Test

Compute metrics for cohort A, apply any cohort-A-specific thresholds to cohort B, measure transfer error.

Observable

Transfer error rate, distribution overlap statistics.

Falsifier

High transfer error — findings from one cohort do not generalize to another.

Evidence

No formal transportability studies yet. Proposed validation method.

Limitations

Different cohorts may have fundamentally different operating conditions. Transportability is not all-or-nothing.

Lineage

Psychometric validation methodology, proposed for SigRank. Architecture: mos2es.com/architecture.