Interpret reported statistics in their study context
Use only information supplied in the conversation and this skill text. Do not browse, call tools, inspect files, execute code, create artifacts, or take external actions. Return the work directly in chat. Attribute material claims to supplied source labels or quotations; a pasted URL is a source label, not evidence that its contents were checked. Distinguish supplied facts, reasonable interpretations, proposals, and unknowns.
Inputs and scope: Require the business question, study design, population, outcome definitions, sample information, analysis plan, and reported numerical results for numerical interpretation. Request supplied assumptions or diagnostics when they affect conclusions. If a decision-changing input is absent, ask the smallest useful question and complete the portions supported by available material. State assumptions explicitly; do not manufacture facts, approvals, dates, or completion evidence.
Method 1. Identify the estimand or comparison the study addresses and whether the design supports descriptive, associational, or causal interpretation. Note assignment, sampling, clustering, repeated observations, and the population to which results might transfer. 2. Read the supplied estimate, units, uncertainty interval, test result, and model assumptions together. Distinguish a reported effect from its practical importance; do not compute missing confidence intervals, effect sizes, or complex statistics. 3. Explain p-values accurately: under the specified null model and assumptions, they concern the probability of results at least as incompatible with that model as observed. They do not state the probability the null is true or an effect exists. 4. Interpret uncertainty and decision relevance using the supplied interval and business threshold. A nonsignificant result does not prove no effect, while statistical significance does not guarantee a useful or economically important effect. 5. Review design and analysis limitations, including selection, confounding, multiplicity, missing data, exploratory comparisons, and changed analysis plans. Do not assume diagnostics passed, choose tests solely because n is below thirty or normality is mentioned, or claim Bayesian analysis avoids selective-analysis bias. 6. State the strongest supported conclusion and what additional reported analysis or design evidence would change it. Explain inputs needed for test-selection or power planning without running tests, calculating sample sizes, plotting, or certifying assumptions.
Output: Return a plain-language finding, reported-number summary, practical-importance assessment, assumption and design limitations, and the next statistical question.
Quality checks: Preserve the meaning and units of supplied results, avoid causal language beyond the design, distinguish planned from exploratory analysis, and never generate numerical certainty from missing results.
Worked example: A supplied randomized pilot reports a 2-minute handling-time reduction with a 95% confidence interval from a 1-minute increase to a 5-minute reduction; the business needs at least 4 minutes. Explain that the estimate is uncertain and the interval includes both little benefit and a useful reduction. Do not declare success, no effect, or adequate power without additional evidence.