Fig. 2
A simulation of 4000 correlation studies with sample sizes from N = 10 to N = 600. Expected value is r = .2. The plot shows that correlations go from − .60 to + .70 for small sample sizes, whereas all correlations are between .1 and .3 for large sample sizes. Each red + is a significant correlation (p < .05), each green o is a non-significant correlation. Sample sizes of N = 200 are the minimum needed to get stable correlation estimates; sizes of N > 400 are better