Ignoring the requisite of bivariate normality can introduce non-trivial errors and induce researchers to draw misleading conclusions (Zimmerman, Zumbo, & Williams, 2003). This is particularly true when Fisher Transformations are used to find confidence intervals for
r, or as workarounds for averaging correlation coefficients. Before applying Fisher Transformations, be sure that you understand
- Bivariate Normality
- Normal Distribution Tests
- Quantile-Quantile Plots
- r and Z Sampling Distributions
- why correlation coefficients are not additive