1

I have to calculate n for my research.

I'm using a calculator (this one: http://www.cct.cuhk.edu.hk/stat/other/correlation.htm). α = 0.05 β = .2 r = ...

I have trouble figuring out what r is. I get that r should be based on effect sizes (in my case correlations) found in earlier research.

But I have many relations in my research and so many correlations to look at.

Should I search the literature for all the relations in my research and go with whichever smallest correlations I find to determine r? Just to be safe?

Hong Ooi
  • 7,629
  • 3
  • 29
  • 52
Stephen
  • 11
  • 1
  • 3
    "value of r, the sample correlation based on N observations", according to the link you provided. It stands for the expected (sample-based or empirical) correlation before seeing any data. In other words, what is the magnitude of the correlation coefficient you expect to find in your data, or according to, e.g., a literature review. If you have multiple outcomes, things might get a bit trickier as you may want to control familywise error or false discovery rate, for example, to avoid the pitfalls of [spurious correlations](http://stats.stackexchange.com/q/5750/930). – chl Oct 01 '14 at 14:22

0 Answers0