Evidence, Assessment, and Remediation

Completion requires executable work, independent reasoning, and a changed case. This rubric is an author-defined standard; it has not been calibrated with learners and makes no claim about reaching a population percentile.

Competency map

ID	Demonstrated ability	Learning evidence	Independent evidence	Repair
PS1	Define sample spaces and observation protocols	A1,A3,A5; worked example1	Quiz1	Enumerate outcomes and selection likelihoods
PS2	Condition with base rates and evidence dependence	A2,A4; Lab A	Quiz2	Rebuild count table; label denominators
PS3	Model stopping and use indicators	B2; Lab A	Quiz3	Draw terminal paths before summing
PS4	Choose distributions from sampling mechanisms	B1,B4,B5; Lab B	Quiz4	Enumerate small samples; compare variances
PS5	Account for covariance	C1,C2; Lab C	Quiz5	Expand variance before dropping terms
PS6	Normalize densities and interpret quantiles	B3; concepts13–14	Quiz6	Check area, units, and CDF
PS7	Separate exact algebra, approximation, and coverage	C3,C4,C6; Labs C–D	Quiz7	Enumerate coverage; compare sampling units
PS8	Make a defensible engineering inference	C5; experiment memo; Lab D	Quiz8	Fix target workload; list missing causal evidence

Pass criteria

1. All six original learner functions and the four inference-extension functions pass their respective public checks; preserve code and reports for two seeds. Explain one failing attempt and its repair. Tests alone do not earn mathematical credit.
2. Submit the model derivations and experiment memo. All five memo criteria must pass.
3. Score at least3/4 on each quiz competency, including its model/assumption point. A strong total cannot compensate for a missing competency.
4. Correct all critical errors: reversed conditioning, unjustified independence, wrong stopping/sampling rule, copied rows counted as independent, averaged quantiles, confidence interpreted as a posterior without a model, or causal conclusions from the aggregate fixture.
5. Complete a changed case in two weak competencies after3–7 days. For each, explain the changed assumption and derive the result unaided. Reassess only the affected competencies after repair.

Record statuses separately: guided, self-assessed, reviewed by another person, delayed recheck complete. If no reviewer is available, proceed after the self-assessment and clearly preserve that limitation. Do not label unreviewed work externally validated.

Transfer prompts

• Compare four retries against four total attempts. Define the model before using a formula.
• Replace independent failures with a shared outage lasting across all attempts. Identify which expectation and success formulas break.
• Sample with replacement from the finite pool. Explain which moments change and which do not.
• Replace independent requests with repeated measurements from ten sessions. Propose an appropriate experimental unit.

Review record

For each competency retain date, evidence path, score, assistance, reviewer, error, corrective action, and recheck result. The instructor guide supplies pilot and review procedures. A genuine learner pilot remains necessary to evaluate workload estimates, clarity, and rubric consistency.

Required extension gates

Complete PS9–PS12 in addition to PS1–PS8, using the same 3/4-per-gate standard including the model/assumption point. Submit the inference memo appendix and both seeded reports. The additional critical errors are omitted transformation support, treating likelihood as a posterior, duplicate-data Bayesian updates, a wrong randomization scheme, discarded ties, or a p-value interpreted as posterior probability. Preserve self-assessment/reviewer status and delayed checks.