Bootstrapped Deattenuated Correlation with Missing Data

Issues with correlation attenuation due to measurement error are well documented. A corresponding correction, the deattenuated correlation, has been known for over a century. For over a decade, researchers have been investigated the deattenuated correlation identifying factors impacting its performance. Nonetheless, the deattenuated correlation is underutilized. In addition, there is limited research concerning confidence intervals for the deattenuated correlation. Here, the bootstrapped deattenuated correlation with corresponding confidence intervals is investigated for simulation conditions not previously considered simultaneously: missing data and non-normal distributions. The bootstrap deattenuated correlation was assessed for relative bias, standard error, and 95% coverage probability for the percentile based and bias-corrected and accelerated confidence intervals. With no missing data, the bootstrap deattenuated correlation performed well under all distributions except for the Pareto.