The fastest way to waste a month is to collect data first and pick the analysis afterwards. These prompts force the order: state what would falsify your belief, choose the control, estimate the sample you need for the effect you care about, and name the test — all before anything is measured.
What's Included
- Null and alternative written so a result could genuinely go either way.
- Control selection with the confound you are guarding against named.
- Sample-size estimate tied to an effect size you would act on.
- The right test for the data shape, decided in advance.
How To Use
- Write down what result would change your mind.
- Generate the design and register it before collecting anything.
- Track the run as records, one per condition.
- Report the result you got, including the boring one.
Frame the question with the research question generator, gather the inputs with survey design prompts, and publish with the research summary report.