Why preregister research?
By preregistering research, you create a public record of your hypotheses and planned analyses before you conduct them, so that you can later demonstrate that you did not change hypotheses after learning your results (HARKing; Kerr, 1998), nor that you changed the analysis to create a specific result. This increases the credibility of your findings, and is increasingly becoming the standard in our field.
Questionable research practices
Preregistration is all about proving that we don't engage in questionable research practices (QRPs).
- Nuzzo (2015) highlights the potential for self-deception in research, and summarizes countermeasures (preregistration being one of them)
- Wagenmakers, Wetzels, Borsboom, van der Mass, and Kievit (2012) list the different biases that can be introduced into an analysis, and argue that only pre-registered research should be labelled 'confirmatory'.
- Simmons, Nelson, and Simonsohn (2011) show how flexibility in the analysis inflates the false positive rate by demonstrating that listening to "When I'm sixty-four" by the Beatles rejuvenates participants.
- John, Loewenstein, and Prelec (2012) estimate the prevalence of questionable research practices such as the above. Fiedler and Schwarz (2016) argue that their estimates were inflated and that the issue is less widespread.
Preregistration as a solution
-
van’t Veer and Giner-Sorolla (2016) provide an excellent overview over the motivation for and history of preregistration in psychology and medicine, discuss advantages, objections, and limitations, and provide practical guidance with a focus on social psychology.
-
In an open letter, Chambers, Munafo and over 80 signatories argue that trust in science would be improved by study pre-registration (2013).
see also
Registered reports are a special case of preregistration, where the study plan is peer-reviewed as part of the publication process. They merit their own entry.
Preregistrations can be augmented or supplanted by alternative methods that may be more practical. Srivastava (2018) describes these strategies for sound inference in complicated research.