A Monte Carlo simulation study compared four bootstrapping procedures in generating confidence intervals for the robust Winsorized and percentage bend correlations. Results revealed the superior resiliency of the robust correlations over r, with neither outperforming the other. Unexpectedly, the bootstrapping procedures achieved roughly equivalent outcomes for each correlation.
King, Jason E.
"Bootstrapping Confidence Intervals For Robust Measures Of Association,"
Journal of Modern Applied Statistical Methods:
2, Article 24.
Available at: http://digitalcommons.wayne.edu/jmasm/vol2/iss2/24