# DEV-SYN-02: Cedar transit survey evidence audit **Family:** source-grounded synthesis. **Version:** 1.0, immutable. **Provenance:** directly hand-authored fictional surveys, records, and claims; not real passenger data or experimental evidence. IDs are local to this task. ## Immutable assignment Write an evidence-audit memo for a fictional transit newsletter editor. Reconcile the corrected survey counts, assess a proposed headline, and specify a cautious replacement headline and two-sentence public summary. Use only S1–S5. Describe survey respondents rather than assuming they represent all riders. Produce a text document, not an executed publication or a model-performance report. ## Public requirements - **R1:** Keep the entire deliverable within 650 words, including a memo, a replacement headline, and a two-sentence public summary. Count whitespace-separated words throughout, excluding punctuation-only table separators. - **R2:** Use the corrected period-2 off-peak count and explicitly identify the earlier count as superseded. - **R3:** Report peak satisfied/respondent counts and percentages for both periods and their percentage-point change. - **R4:** Report off-peak satisfied/respondent counts and percentages for both periods and their percentage-point change. - **R5:** Pool counts within each period, report overall percentages to one decimal place, and report the overall percentage-point change to one decimal place. Do not average the two group percentages without weighting. - **R6:** Calculate response rates among invited riders for each period; keep that denominator distinct from respondents used in satisfaction rates. - **R7:** Explain how the peak/off-peak composition of respondents changes; do not interpret the overall increase as an increase in every group. - **R8:** Reject the proposed headline's claim about every group and its causal claim about the timetable, using packet evidence. - **R9:** State that respondents are self-selected, the periods do not track the same people, and group-specific invitation counts are unavailable. Do not invent subgroup response rates or population satisfaction. - **R10:** Supply a replacement headline and two-sentence summary consistent with the corrected counts, mixed group changes, and lack of causal evidence. No particular wording is required. - **R11:** Identify two useful follow-up information needs without claiming the information has been collected or that another survey has occurred. - **R12:** Cite source IDs and line locations for factual claims and calculation inputs, use only S1–S5, and label the supplied surveys as fictional rather than real research outcomes. ## Complete authorized source packet ### S1 — Period-1 respondent table - **L1:** Period 1 has 100 peak respondents, of whom 90 marked satisfied; it has 50 off-peak respondents, of whom 20 marked satisfied. - **L2:** Within a period, the groups are disjoint and cover all respondents. Each respondent supplied exactly one satisfaction response. ### S2 — Period-2 correction - **L1:** Period 2 has 60 peak respondents, of whom 48 marked satisfied, and 90 off-peak respondents. - **L2:** The signed correction gives 63 satisfied off-peak respondents in period 2, explicitly replacing the earlier draft's 54. No other count changes. ### S3 — Invitations and sampling - **L1:** Periods 1 and 2 invited 375 and 250 distinct riders respectively. Each period received the respondent counts in S1 or S2; there are no other or missing response categories. - **L2:** Participation was voluntary. No sampling weights or group-specific invitation counts are supplied. The two periods do not follow the same individuals; invitation records cannot establish satisfaction among nonrespondents. ### S4 — Proposed publication claim - **L1:** A draft headline reads: “New timetable raises satisfaction in every rider group.” - **L2:** A timetable change occurred between the periods in this fictional scenario. There is no randomized assignment, comparison service, or evidence separating timetable effects from who responded or other changes. ### S5 — Editorial scope - **L1:** The editor wants corrected descriptive evidence, not a declaration that service quality or population satisfaction improved. - **L2:** Follow-up possibilities include learning why invitation totals or respondent composition changed, collecting group-specific invitation counts, or designing a comparable sample. These are information needs, not completed actions or guarantees of causal identification.