Data-Driven and Original-Research Blog Posts

Apply ../00-blog-writing-guardrails/SKILL.md first; its accuracy, claims-substantiation, and E-E-A-T gates apply to every statistic and claim in this skill. The gates below add requirements specific to posts built around survey data, original research, or third-party datasets.

1. Methodology-transparency gate

• Before publishing any post built on a survey or study, disclose how the data was collected: the survey mode, the population studied, how the sample was recruited or selected, and the exact question wording behind any reported figure. [AAPOR-DISCLOSURE-01]
• State whether the sample was drawn using probability-based selection or a non-probability method (opt-in panel, convenience sample, scraped dataset); do not present a non-probability sample's results with the same confidence language used for a probability sample. [AAPOR-DISCLOSURE-01]
• Be transparent about the assumptions, methods, limitations, and possible sources of error behind any statistic before publishing it, not only the headline number. [ASA-ETHICS-01]

2. Sample-size and margin-of-error disclosure gate

• Report the sample size for every reported statistic, broken out by subgroup when a subgroup figure is cited; a subgroup number with an unstated (and likely small) base is not publication-ready. [AAPOR-DISCLOSURE-01]
• For probability-based samples, disclose the estimated margin of sampling error alongside the headline statistic; if a precision estimate is given for a non-probability sample, describe the model that produced it rather than presenting it as a standard margin of error. [AAPOR-DISCLOSURE-01]

3. Anti-p-hacking and anti-cherry-picking gate

• Disclose when multiple comparisons, subgroup cuts, or alternative time windows were tested on the same dataset, and note any adjustment made for that multiplicity; do not report only the one cut that reached significance while silently dropping the others. [ASA-ETHICS-01]
• Resist selectively interpreting or reporting data to fit a predetermined narrative; if the full dataset does not support the post's thesis, revise the thesis rather than the presented slice of data. [ASA-ETHICS-01]