If you can address these TEN items you will have built the engine of a Stage 1 Registered Report

1) What is the main question being addressed in your study?

• Why is it important that we answer this question? What’s the big picture?

2) Describe the key independent and dependent variable(s), specifying how they will be measured.

• Registered reports are based on specific and precisely defined manipulations and measurements driven by clear hypotheses. Make sure the reviewers can clearly connect your hypotheses to your experimental manipulations and that it is very clear what is being measured and how the measurements will be carried out and quantified.

3) What are your hypotheses?

• All registered reports must be hypothesis driven, and the hypotheses must be specifically about the relationships among your independent and dependent variables, as operationalized by the measurements you will be making.
• It is often useful to list each hypothesis separately, as Hypothesis 1, Hypothesis 2 etc. and to pair each with a specific null hypothesis, if that is the statistical approach being used. 

4) How many and which conditions will participants/samples be assigned to?

• Where applicable be sure to include details of randomization, anonymizing, and counterbalancing. Describe the design very precisely, especially in terms of comparisons. Will data comparisons be within-subjects, between-subjects, mixed, or something else?

5) How many observations will be collected and what rule will you use to terminate data collection?

• Ensure that your stopping rule takes into account any data exclusions.
• If adopting null hypothesis significance testing, what power will your study achieve? What effect size will you target and why? Remember that you are choosing the smallest effect size of theoretical or applied interest, or the smallest you can feasibly detect. For an actual RR you can use pilot data to help motivate this estimate, but you shouldn’t rely on pilot data alone because it is vulnerable to bias.
• If adopting Bayesian sampling methods, make sure to clearly describe your prior distribution assumptions. Similarly, if using a Bayes factor for asserting relative support of H0 or H1, what is the value that is needed to support your H1?

6) What are your study inclusion criteria?

• How will participants/samples be recruited/included and under what specific rules?

7) What are your data exclusion criteria?

• State rules for excluding data both at the level of samples/participants (within groups) and at the level of raw data (within samples/participants), e.g. conditions involving data quality, completeness and outliers.
• Remember to be comprehensive: exclusion criteria are very difficult to change after data collection has commenced because doing so risks introducing bias. Think about previous experiments you have done and all the reasons you have ever thrown out a data set or data point.

8) What positive controls or quality checks will confirm that the obtained results are able to provide a fair test of the stated hypothesis?

• WHAT’S THIS? A positive control tests the existence of phenomena that would confirm that the independent variables, dependent variables, or instrumentation was used correctly and is therefore capable of testing the main study predictions. One of the most famous positive control experiments was the use of the Galileo spacecraft to test for the existence of life on Earth. If the instrumentation on the probe couldn’t detect life on Earth (i.e. had the positive control failed), then it would not be reasonable to use to the probe to test the hypothesis that life existed on other planets.
• Not all experimental designs have suitable positive controls. Where a positive control isn’t possible, think of what quality checks or verifications you would build into your design before results are known to convince a skeptic that you had conducted the experiment to a sufficient standard (e.g. noise within certain limits etc.). Make sure these are independent of your main hypothesis tests.
• Where a positive control (e.g. manipulation check) or quality check (e.g. lack of floor or ceiling effects in data) requires a statistical test, ensure that the test is adequately powered or sampled. 

9) Specify exactly which analyses you will conduct to examine the main question/hypothesis(es)

• Ensure that there is an exact correspondence between each scientific hypothesis and each statistical test. Failure to precisely specify these links is one of the main reasons RRs are rejected.
• If your analysis strategy will depend on the results (e.g. normal vs. non-normal distribution) then specify the contingencies for making different choices, i.e. IF-THEN statements.
• In the event of a negative result, would you be happy to conclude that there “was no evidence of a difference” between conditions, or would you instead want to be able to make the stronger claim that “there is evidence of no difference between conditions”? The first inference is limited to absence of evidence while the second (stronger) one refers to evidence of absence. If you want to make the stronger inference, you will need Bayesian inferential methods or frequentist equivalence testing.
• Complete the design planner below to make the links absolutely clear between the research question (or questions), hypothesis (or hypotheses), sampling plans, analysis plans, and contingent interpretation