a-priori power analyses and reporting effect sizes

[Daniel Lakens]

After trying to perform a-priori power analyses for my replication study for the reproducibility project, I was slightly annoyed with myself that I really didn't understand much about it. While trying to figure out how to correclty report effect sizes and perform power analyses, I wrote a paper about it. 

In this blog post (The first rule of not understanding effect sizes is you don’t talk about not understanding effect sizes) I link to the manuscript and a spreadsheet to help report effect sizes correctly: https://sites.google.com/site/lakens2/blog/thefirstruleofnotunderstandingeffectsizesisyoudon’ttalkaboutnotunderstandingeffectsizes

Some useful things for this group are that I explain how to calculate partial eta squared from the F-value and degrees of freedom (so you can always do a power analysis, even if no means, SD's, n's or correlations are reported, both for within as for between designs). Also, I explain why for within designs, you should not insert the partial eta squared you got from SPSS directly into G*Power. By default, this will provide the wrong sample size estimate. Instead you should find the option box, and check 'as in SPSS' (only available in the latest version of G*Power).

Another idea you might find interesting in the manuscript is reporting the sample size needed for a replication, when you report your study (I call it Nrep). It might be a way to provide a preliminary (although not unbiased) way to judge the power of the performed study. The APA manual asks that we discuss the power of a study, but we never do, because there are no recommendations of how to do this (and post-hoc power analysis are obviously wrong). 

Hope this is helpful.

Daniel

[Gustav Nilsonne]

Thank you Daniel!

I have been struggling a lot with this while coding articles and trying to perform power calculations. Your explanations will be very helpful.

Best wishes, Gustav

Gustav Nilsonne, MD, PhD
Postdoc

+46 (0) 736-798 743

Stockholm University 

Stress Research Institute

106 91 Stockholm

gustav....@stressforskning.su.se

Karolinska Institutet 

Department of Clinical Neuroscience and Osher Center for Integrative Medicine
Retzius väg 8 A2:3
171 77 Stockholm

gustav....@ki.se

---

From: openscienc...@googlegroups.com [openscienc...@googlegroups.com] on behalf of Daniel Lakens [lak...@gmail.com]
Sent: 15 May 2013 10:13
To: openscienc...@googlegroups.com
Subject: [OpenScienceFramework] a-priori power analyses and reporting effect sizes

--
You received this message because you are subscribed to the Google Groups "Open Science Framework" group.
To unsubscribe from this group and stop receiving emails from it, send an email to openscienceframe...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
 
 

[Roger Giner-Sorolla]

This would have been really useful - still is - for figuring out effect size and power for within-designs in the Reproducibility Project.