Disattenuation of Correlations Due to Fallible Measurement

The downward bias in correlation coefficients due to measurement error has long been recognized in the quantitative methodology literature. Whether to adjust for such bias in attenuated correlations, however, remains a heated debate among methodologists. As a result, applied researchers have largely ignored the controversial practice of treating observed correlations as error free. This article revisits the theoretical framework for disattenuation of correlations and critically examines the current reporting practice in applied research. I begin by summarizing the rationale for disattenuation and obstacles in implementing such disattenuation. I then highlight the chaotic status in applied research with a recent meta-analysis study on neuroimaging research. I conclude by calling for a renewed interest in addressing this disconnect between theory and practice and providing some practical guidelines for researchers and practitioners.

Keywords: Disattenuation, Correlation, Measurement error, Unreliability