Useful Resources

Effect size calculators (the best online tools for calculating d or r)
Effect size equations (the secrets revealed!)
The Result Whacker (controversial but fun!)
Thresholds for interpreting effect sizes (what does your result mean?)
References (a good reading list for learners)

Three how-to books…

7 Responses to Useful Resources

  1. Joseph Green says:

    Dear Professor Ellis,

    I would like to compute an effect size for the difference between the areas under two receiver-operating-characteristic curves.

    There are methods for computing standard errors, variances, and covariances of estimates of those areas, and of course p values, but I haven’t seen an effect size index.

    Do you know of any?

    J Green

    • Paul Ellis says:

      Hi Joseph,
      What field are you in? I am not familiar with “receiver-operating-characteristic curves” so I’m afraid I can’t suggest an ES index. If you are dealing with samples, and it sounds like you are,then I am sure you must be able to calculate an effect size index. Usually the best way to find out the names of indexes relevant to your particular discipline is to Google “effect size” along with terms associated with your area. You’d be surprised what turns up. All the best with your research.

  2. Anna says:

    Hi Joseph,

    the area under curve already is a kind of effect size. There is a short paper by Rice and Harris from 2005 (DOI: 10.1007/s10979-005-6832-7) where you can find a table comparing AUC, d, and r, to by able to compare these indexes. This is based on tables by Pearson and Hartley, 1954.


  3. Esther Tschetter says:

    I bought The Essential Guide to Effect Sizes and your pdf on Effect Size Matters. On page 99 in The Essential Guide to Effect Sizes you state that “Before these effects can be combined they will need to be transformed into a common metric”. Is there a way or book that will show show me how to turn a R squared and Beta into a Pearson r?
    Thank you

    • Paul Ellis says:

      To turn an R-squared into an r you unsquare it. Actually, that’s not technically a smart thing to do because an R-squared will typically reflect the associated between numerous variables while an r measures the degree of association between two.

  4. Esther Tschetter says:

    Thanks for your help. Happy Holiday!

  5. MaryAnne Haines says:

    Dear Professor Ellis
    I have just come across your book ‘The Essential Guide to effect sizes’, which is an excellent guide! I have conducted an exploratory study in educational research with a small sample and have used non parametric statistics for the ranked data re Spearman rho and Kruskal Wallis (and follow up analyses – Pairwise comparison). For non parametric tests using the effect size from the r family would I need to report on the confidence interval? Could you recommend a text for a beginner re effect sizes and their interpretation? Is it acceptable calculating effect sizes for small samples (e.g; comparing three groups of approximately 10)
    Thank you

%d bloggers like this: