Replication Case Sampling Cycle¶

Replication process — instantiates Structured Comparative Case Design

Adds new cases in deliberate rounds — some expected to repeat the result, some expected to overturn it — to map where a finding holds and where it stops.

Replication Case Sampling Cycle grows the case set on purpose, round by round, treating each new case like an experimental replication with a predicted result. A literal replication is a case chosen because the theory says it should show the same outcome; a theoretical replication is a case chosen because the theory predicts a different outcome, for a stated reason. Confirmation comes not from piling up similar cases but from the pattern holding: literals repeat, theoreticals diverge as forecast. What makes it THIS mechanism is that it is proactive and generative — it manufactures new evidence to extend and bound a finding, rather than reasoning over a fixed set.

Example¶

A small coastal fishery adopts a community catch-quota rule and, within two seasons, stocks and incomes both recover. Encouraging — but one case. The replication cycle plans the next rounds. Literal replications come first: three more small, socially cohesive fisheries where the theory says the rule should work again; it does in two and, tellingly, fails in the third. Then a theoretical replication: a large, anonymous industrial fishery where the theory predicts the rule should fail, because enforcement depends on the social cohesion the big fishery lacks — and it does fail, as forecast. The failed literal is handed to a follow-up investigation; the predicted failure of the theoretical case is a win, sharpening the boundary. After a few rounds the finding is no longer "quotas work" but "community quotas work where the community can watch itself," with the edge of that claim mapped by cases on both sides of it.

How it works¶

The cycle iterates: state a prediction for the next case, select a case that tests it (literal to confirm, theoretical to probe a boundary), run it, and update — a confirmed prediction extends the finding, a failed literal forces revision, a confirmed theoretical failure sharpens a scope condition. Its distinguishing feature is predicted-result sampling: each case earns its place by the specific expectation it tests, and the cycle stops when new cases stop changing the boundary (theoretical saturation), not at a case count.

Tuning parameters¶

Literal / theoretical mix — how many confirming versus boundary-probing cases per round. More literals build confidence; more theoreticals map limits faster.
Round size — how many cases before pausing to re-read the pattern. Larger rounds are efficient but risk over-committing before a failed replication is noticed.
Stopping rule — when added cases stop moving the boundary. A strict rule keeps sampling; a loose one may quit before the edge is found.
Failed-literal handling — whether a case that should have replicated but did not triggers theory revision, a measurement re-check, or a hand-off to deviant follow-up.

When it helps, and when it misleads¶

Its strength is that it builds genuine external validity case by case and draws the map of where a finding applies — and a failed literal replication is among the most informative results a study can get. Its costs are real: the cycle is slow and expensive, and "literal" replications chosen too similar to the original teach almost nothing.^[1] The classic misuse is the file-drawer in miniature — counting the confirming replications and quietly setting aside the ones that failed, which inverts the whole point. The discipline that guards against it is to state each case's predicted result before running it and to report the failed replications as prominently as the successes.

How it implements the components¶

case_selection_rationale — here selection follows replication logic: each new case is chosen as a literal or theoretical replication of a stated prediction.
inference_scope_and_boundary_conditions — the accumulating pattern of confirmed and forecast-to-fail cases is what maps how far the finding travels and where it stops.

It does not build a single matched comparison pair — that is Matched Case Pairing Protocol; it does not perturb a fixed case set to test fragility — that is Sensitivity to Case-Set Analysis; and it does not run the reactive re-investigation of a specific anomaly — that is Deviant Case Follow-Up Protocol.

Instantiates: Structured Comparative Case Design — the cycle is how the design earns and bounds generalization beyond its first cases.
Consumes: Case Universe Sampling Frame supplies the bounded pool from which literal and theoretical replications are drawn.
Sibling mechanisms: Sensitivity to Case-Set Analysis · Deviant Case Follow-Up Protocol · Most-Similar Systems Design · Most-Different Systems Design · Matched Case Pairing Protocol · Measurement Equivalence Audit · Within-Case Process Tracing · Rival Explanation Elimination Table · Case Selection Bias Audit · Case Universe Sampling Frame · Comparative Case Review Panel · Comparative Historical Timeline · Configurational Comparison Truth Table · Counterfactual Contrast Memo · Cross-Case Evidence Matrix Tool

Notes¶

Distinguish it from Sensitivity to Case-Set Analysis: replication builds outward, adding real new cases to extend and bound a finding, while sensitivity works inward, re-running the analysis on the cases already in hand to see whether the conclusion is fragile. They answer different questions — "how far does it go?" versus "does it even hold here?" — and are strongest used together.

References¶

[1] The distinction between literal replication (cases predicted to reproduce a result) and theoretical replication (cases predicted to produce a contrasting result for a stated reason), and the "replication logic" that treats a multi-case study as a series of experiments rather than a sample (Yin). Confirmation rests on predictions holding across both kinds, not on case count. ↩