# Preregistration (https://terac.com/docs/researchers/research-methods/preregistration)

Lock in your hypotheses and analysis plan before you collect data to guard against p-hacking and HARKing.

Preregistration means documenting your research plan, your hypotheses, your design, and your analysis, and timestamping it in a public archive **before** you collect or look at your data. It is the strongest single thing you can do to make a confirmatory study credible.

## Why Preregister [#why-preregister]

When the analysis plan is fixed in advance, you remove the researcher degrees of freedom that quietly inflate false positives:

* **p-hacking:** trying many analyses and reporting only the one that crossed significance.
* **HARKing:** presenting a post hoc hypothesis as an a priori one. See [Research Design](https://terac.com/docs/researchers/research-methods/research-design).
* **Selective reporting:** dropping conditions, outcomes, or exclusions that did not work out.

A preregistration lets a reader distinguish what you predicted from what you discovered. Both are valuable, but only the first counts as confirmatory evidence.

## What to Include [#what-to-include]

A useful preregistration is specific enough that someone else could run your analysis without asking you anything:

| Section        | What to specify                                                               |
| -------------- | ----------------------------------------------------------------------------- |
| **Hypotheses** | The exact predictions, stated directionally where possible                    |
| **Design**     | Conditions, manipulations, and how participants are assigned                  |
| **Sample**     | Target size, how you arrived at it, and stopping rule                         |
| **Measures**   | Every variable and exactly how it is computed                                 |
| **Exclusions** | Rules for removing participants or responses, written before you see the data |
| **Analysis**   | The specific test for each hypothesis, including how you handle covariates    |

> Decide your participant exclusion rules in advance. Removing data after seeing
> results, even for defensible reasons, reopens the door to bias.

## Where to Preregister [#where-to-preregister]

Independent registries timestamp your plan so it cannot be edited silently:

* **OSF (Open Science Framework):** flexible, widely used, supports embargoes.
* **AsPredicted:** a short standardized form, good for simple designs.

## Registered Reports [#registered-reports]

A **Registered Report** goes one step further. You submit your introduction, methods, and analysis plan to a journal for peer review *before* collecting data. If the plan is sound, the journal grants **in-principle acceptance**, committing to publish the results regardless of how they turn out. This removes publication bias toward positive findings and rewards good design over lucky results.

## Preregistration and Terac [#preregistration-and-terac]

A few platform settings are worth fixing in your preregistration because they cannot change once an opportunity is live:

* Your [participant cap](https://terac.com/docs/researchers/opportunities/creating-an-opportunity) and any [quotas](https://terac.com/docs/researchers/opportunities/screening).
* Your [filters](https://terac.com/docs/researchers/recruitment/filters) and screening criteria.
* Your attention and quality checks, which double as preregistered exclusion criteria.

## What's Next? [#whats-next]

- [Sampling and Power](https://terac.com/docs/researchers/research-methods/sampling-and-power): Justify your target sample size before you register it
- [Analysis and Inference](https://terac.com/docs/researchers/research-methods/analysis-and-inference): Choose the analysis you will commit to