Puckworks · an analysis-autopsy story

How We Falsified Our Own Espresso Headline

The project correcting its own aggregation and interpretation — not a verdict on the source authors.

OBSERVED evidence (corrected observed result): independent

Loading the generated data…

Summary (JavaScript is off)

This page normally lets you walk through four scenes: the superseded headline, the audit, the corrected data, and what a model can and cannot claim. With JavaScript disabled, the complete result — every number and caveat — is in the generated static text summary and the summary image below. You can also download the generated data.

A scatter chart of observed espresso extraction yield in percent against grinder dial. Three dial cells (dial 1.4, dial 1.7, dial 2.0) each show individual replicate points as small squares with a horizontal line at the cell mean. The finer cell sits lowest, the middle cell sits just below the coarse cell, and a label reads the dial 1.7 minus dial 2.0 contrast as a small negative extraction-yield value. The exact numbers are in the static text summary linked above.

Walk through the revision

Choose a scene

·

1 · The headline we wanted SUPERSEDED

The attractive first reading was an interior “fine-grind” extraction-yield maximum — a peak at the middle dial. It came from a literal evaluation of the rounded published response- surface coefficients at the central condition (about · g, treated as if it were the real level) together with a target that mixed milligram-scale solute masses with gram-scale TDS. We are not hiding it; here is the whole revision, in order.

    We keep the earlier interpretation visible. This is the project correcting its own aggregation and interpretation — not a verdict on the source authors.

    Each panel carries its own evidence label

    The global OBSERVED badge refers to the corrected observed replicate result only. The other panels are labelled separately:

    What this does not prove

    Scope. ·

    Caveat. ·

    Fidelity ceiling. ·