[Library](https://stochastly.com/library)

# Research method

Rules of method for quantitative research, each stated with the error it prevents and how to apply it.

## What this shelf covers

Short rules for running a test that can come out negative. Each note names the error it prevents and the check that applies it: register the hypothesis and the decision rule before the data are read, validate the instrument with a positive control of known size, build the null from the same sample, and report the number of trials with the result. A rule here is a procedure with a measurable output, such as a detection floor in the units of the effect or a count of configurations tried.

## 3 notes

[A backtest is reported with the number of trials, or it is not reported](https://stochastly.com/library/research-method/a-backtest-is-reported-with-its-trials)

[Naive multiple-testing corrections manufacture false negatives on correlated candidates](https://stochastly.com/library/research-method/correlated-candidates-break-naive-corrections)

[The practices that make a result recomputable are decided before the research](https://stochastly.com/library/research-method/the-practices-that-make-a-result-recomputable-are-decided-first)

---
Source: https://stochastly.com/library/research-method
