Likelihood-Ratio Frame¶
Reasoning tool — instantiates Conditioned Probability Frame Specification
Separates the strength of the evidence from the probability of the hypothesis by expressing what a signal says as a ratio that updates a base rate rather than replaces it.
The deepest confusion in conditional probability is mistaking how strong a piece of evidence is for how likely the hypothesis now is. Likelihood-Ratio Frame is the reasoning tool that keeps those two apart. It expresses the evidence as a likelihood ratio — how much more probable the observed signal is when the hypothesis is true than when it is false — and treats that ratio as a multiplier on a base rate, never as the answer itself. Its defining move is factoring the problem: evidence strength (the ratio) times prior prevalence (the base rate) yields the posterior. A powerful signal against a rare condition can still leave the condition unlikely, and this frame is built to show exactly that.
Example¶
A screening test for a disease that affects 1 in 1,000 people comes back positive. The test is "95% accurate," which everyone hears as "95% chance I have it." The Likelihood-Ratio Frame refuses the shortcut. It asks two questions the accuracy figure blurs together: how much more often does a positive appear in the sick than the well? Say the sensitivity is 95% and the false-positive rate is 5% — a likelihood ratio of 19. Then it multiplies that ratio against the base rate, which is 1 in 1,000. Nineteen times a tiny prior is still a small number: the post-test probability of disease lands near 2%, not 95%. The evidence was genuinely strong (a ratio of 19 is not nothing), yet the hypothesis remains unlikely, because the base rate was doing most of the work all along.
The value is a corrected read on a positive result — and, crucially, a frame that says why the intuitive 95% was wrong: it silently discarded the prevalence the ratio was supposed to update.
How it works¶
- Compute the ratio, not the raw conditional. Express the evidence as P(signal | hypothesis) ÷ P(signal | not-hypothesis). This isolates evidential strength from prevalence.
- Anchor on the base rate. State the prior prevalence explicitly as the thing the evidence updates; it never leaves the calculation.
- Update by multiplication (in odds). Convert the prior to odds, multiply by the likelihood ratio, convert back — the renormalization that turns prior into posterior.
- Label the output as posterior. The result is P(hypothesis | signal), tagged so it is never re-read as the evidence strength that produced it.
Tuning parameters¶
- Ratio granularity — a single lumped likelihood ratio versus separate ratios for graded evidence levels (weak / moderate / strong positive). Graded ratios extract more signal but demand more data to estimate.
- Base-rate source — population prevalence versus a subgroup-specific prior. A tighter reference class sharpens the posterior but shrinks the sample the prior rests on.
- Odds vs. probability presentation — carrying the update in odds (where it's a clean multiply) versus probabilities (more intuitive, messier math). Odds are robust; probabilities communicate.
- Number of updates chained — one signal versus several multiplied in sequence. Chaining is powerful but assumes the signals are conditionally independent, which is often false.
When it helps, and when it misleads¶
Its strength is that it dissolves the most stubborn error in the whole archetype — the base rate fallacy, in which a strong-looking test result is read as a strong-looking diagnosis while the prevalence is ignored[1]. By making evidence strength a ratio rather than an answer, it forces the base rate to stay in the room.
Its failure mode is chaining ratios that aren't independent: multiply two correlated signals as if each brought fresh information and the posterior is badly overstated. The classic misuse is treating the likelihood ratio as if it were the posterior probability — quoting "the evidence is 19-to-1" as though the hypothesis were 19-to-1, which is the very confusion the tool exists to prevent, now dressed in its own vocabulary. A second trap is reading the update causally: a signal that raises the probability of a condition is not thereby a cause of it. The guarding discipline is to state the base rate out loud every time, check independence before chaining, and label the output as a posterior probability, not an effect.
How it implements the components¶
base_rate_reference— the prior prevalence is a named, non-negotiable term the evidence multiplies; the frame is built around keeping it visible.renormalization_rule— the odds-multiply-and-convert-back step is the explicit rule that recomputes probability inside the post-evidence frame.probability_type_label— it sharply distinguishes likelihood, prior, and posterior, labeling each so the reader never conflates evidence strength with hypothesis probability.causal_nonclaim_warning— it flags that an updated probability is associational, not an intervention effect.
It does NOT implement joint_probability_table — laying out the four joint cells to expose P(A|B)-vs-P(B|A) inversion is the two-by-two probability table's job; the ratio frame reduces the evidence to a single multiplier instead of tabulating cells. Nor does it implement conditioning_context_statement as a communication act — that is the given-that clause.
Related¶
- Instantiates: Conditioned Probability Frame Specification — this tool distinguishes evidence likelihood from posterior probability within a declared frame.
- Consumes: Reference Population Note supplies the base rate the ratio updates.
- Sibling mechanisms: Conditional Probability Annotation · Frame Compatibility Review · Given-That Clause · Probability Tree · Reference Population Note · Scenario Condition Card · Stratified Rate Table · Two-by-Two Probability Table
Editorial Notes¶
Form Classification¶
Form family: Analysis, Modeling & Optimization
Rationale: Likelihood-Ratio Frame operates as a computation, comparison, model, or analytic representation used to infer, estimate, or choose because it separates the strength of the evidence from the probability of the hypothesis by expressing what a signal says as a ratio that updates a base rate rather than replaces it.
Independent corroboration: The frozen evidence defines Likelihood-Ratio Frame as 'Separates the strength of the evidence from the probability of the hypothesis by expressing what a signal says as a ratio that updates a base rate rather than replaces it', so its operative form is Analysis, Modeling & Optimization.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Statistics & Experimental Design
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Universal
Rationale: The frame derives from statistical likelihood and Bayesian updating, where evidence modifies rather than replaces prior odds.
Related originating lineages:
- Philosophy — Bayesian epistemology materially shaped the separation of evidential strength from posterior belief and base rates.
Review resolution: Both independent reviews assign primary provenance to statistics_experimental_design. The queued secondary differences (alternate_origin_disagreement, origin_mode_disagreement) are reconciled by retaining philosophy only as formative or independently established lineage(s), not merely as application domains. origin_mode=cross_disciplinary_synthesis records the provenance relationship, while domain_reach=universal separately records applicability breadth. confidence=high preserves the more cautious assessment, and encyclopedia_synthesis=false records whether either reviewer identified a corpus-specific synthesis.
Review outcome: Reconciled after independent review; high confidence.
References¶
[1] Bar-Hillel, M. "The Base-Rate Fallacy in Probability Judgments". Acta Psychologica 44(3), 211–233 (1980). Describes the base-rate fallacy as overweighting case-specific diagnostic information while neglecting relevant base rates. registry ↩