PROGRAM

12:00 pm: Talk by Daniël Lakens

Sample Size Justification: Designing Well-Powered but Efficient Studies

Abstract: If there is one issue researchers agree on in the current ‘crisis of confidence’ in psychology, it’s that we need to collect data larger samples to achieve high statistical power. Many journals now explicitly require researchers to justify their sample sizes. In this talk, I will discuss how to design well- powered studies with sequential analyses, and when to pre-register one-sided tests to improve the efficiency of data collection. I will highlight some common pitfalls when performing power analyses, and how power analyses (both on the published literature, as on your own studies) are always bia- sed estimates of the required sample size. Finally, I will explain why even when you design high-powered lines of research, non-significant effects should be expected, and explain when sets of studies with mixed results should be interpreted as evidence for a true effect.

– Should we all become Bayesians?

– How can I evaluate a literature riddled with false positives?

– Why isn’t 95% of published findings true, if we all use an alpha level of 0.05?

– How can I provide support for a null effect?

– Is it true that the null is never true?

We strongly invite you to send us your own questions about data analysis (at oklein@ulb.ac.be) one week before the seminar at the latest 

———–

** Registration to the event is free but mandatory via this online form.

[Details about the exact location of the seminar will be determined shortly, once the amount of participants is known]