Addressing Missing Data in Quantitative Counseling Research
Ryan M. Cook
Abstract
Although many statistical procedures that are utilized by counseling researchers require complete datasets, the problem of missing data represents a common analysis hurdle that must be overcome. Typically, counseling researchers address the problem of missing data using listwise deletion; however, this procedure has some statistical disadvantages (e.g., unnecessary reduction in statistical power and unintentional introduction of bias). The most recent versions of statistical packages such as SPSS now include more robust imputation procedures for dealing with missing data. However, utilizing any deletion or imputation procedures without a thorough understanding of the conditions in which these procedures should be used could negatively impact study findings. In this article, strategies for detecting missingness mechanisms and appropriately handling missing data using deletion and imputation available procedures in SPSS are discussed. The specific procedures reviewed include listwise deletion, pairwise deletion, mean substitution, expectation-maximization, hot deck, multiple imputation linear regression, and predictive mean matching.