Research Navigator

The full study loop, handed to Claude

A reviewer read the headline figure, circled it, and asked a plain question: show me where this number comes from. The author opened the analysis folder and found three versions of the extraction script, a notebook edited live during the writing, and a results file whose timestamp predated the final code. That figure had been real on the day it was generated. Reconstructing it a month later meant guessing which combination of script, parameters, and input files had produced it — and the guesses disagreed by enough to change the conclusion.

Research collapses when the steps drift out of agreement or when the trail from a stated result to its source data goes cold. This book builds the trail, step by step, so the answer to where does this number come from is always a file, a version, and a command anyone can run.

Ten chapters that walk the loop end to end

• research-has-a-method — The shape of an empirical study and why it repeats regardless of subject: survey, isolate, design, measure, compare, verify, record.
• mapping-the-domain — A frontier map of a field’s open questions, structured so the gap you intend to fill is visible before the first data point is taken.
• finding-your-gap — A gap analysis that states your contribution precisely, in terms the field already uses, so scope is set before any measurement begins.
• designing-a-methodology — A methodology spec with a coding scheme, written precisely enough to execute — so there is no ambiguity about what was measured or how.
• building-the-measurement-pipeline — From raw signal to feature table, with the heavy compute offloaded to a remote machine so the laptop stays free during long extraction runs.
• comparison-and-confounds — Before vs. after, with confounds named ahead of time and effect sizes reported alongside significance — because a p-value without a magnitude is an incomplete finding.
• reproducible-findings — Attaching an uncertainty figure to every stated number, so each claim knows what it is claiming and what would falsify it.
• verifying-every-claim — The verifier that recomputes each stated result from source data and keeps only the ones that match — the step that earns the right to publish.
• adversarial-review — A pass that hunts overstated claims before release, so reviewers find nothing the author missed and every surviving claim is one that tried to die.
• the-research-record — Data references, methodology version, code, and the verification report bundled into something a stranger can rerun — the artifact that outlasts the session.

A preview

A reviewer read the headline figure, circled it, and asked a plain question: show me where this number comes from. The author opened the analysis folder and found three versions of the extraction script, a notebook that had been edited live during the writing, and a results file whose timestamp predated the final code. That figure had been real on the day it was generated. Reconstructing it a month later meant guessing which combination of script, parameters, and input files had produced it, and the guesses disagreed by enough to change the conclusion. The finding was withdrawn. Its data had been sound; the path from data to claim had been paved over as the work moved forward.

That verifier is the spine of the whole method. Every other step produces a claim; this is the step that earns the right to publish it, by recomputing the number from the data and keeping only the ones that match. A study built this way survives the question that ended the one above, because the answer to where does this come from is a file, a version, and a command anyone can run.

Researchers who want the loop to close

Anyone running structured empirical work — analysts, operators, independent researchers — who has arrived at the end of a study and found that a key result could not be cleanly reconstructed. The method in this book works for any subject where the study shape is: survey, isolate, design, measure, compare, verify, record. The worked example is an acoustic measurement study; the pattern transfers without modification. The measurement pipeline connects naturally to the data certification work in Building Data QA Infrastructure (Book 17), and the adversarial review shares the gate-and-verify discipline from Self-Verifying Pipelines (Book 7).

A longer excerpt is available to newsletter subscribers.

Buy Ebook on Amazon  Buy Paperback on Amazon