What problem does it solve? Descriptive statistics often look clean while hiding silent errors: composition shifts, missing deflation, wrong denominators, or heavy-tailed means that describe no one. This Skill applies a rigorous discipline to descriptive work so trends, Table 1 summaries, and stylized facts survive composition checks, robustness re-cuts, and external plausibility anchors before they reach a paper or deck. ## Core Features & Use Cases - Comparability enforcement: Fixes denominator, deflation base year, per-capita scaling, weighting, unit of observation, window, and aggregation level before any plot is made. - Composition checking: Decomposes aggregate changes into within-group vs. mix effects (shift-share, Oaxaca, subgroup plots) to catch Simpson's paradox and selection artifacts. - Robustness and validity gates: Re-cuts reported facts across alternative specifications and triangulates levels against external benchmarks or known shocks. - Causal firewall: Enforces descriptive verbs only, routing causal questions to the appropriate identification workflow. - Use Case: You have a hospital panel from 2008-2022 and want a chart of average charges over time. The Skill deflates to constant dollars, checks whether hospital entry/exit drives the trend, reports medians alongside means for the heavy-tailed charge variable, and keeps the write-up descriptive. ## Quick Start Ask the assistant to show the trend in a variable, build a Table 1 by group, or produce stylized facts from your dataset, and it will apply the comparability, composition, and robustness checks before delivering the exhibit.
