Treating missing data in a clinical neuropsychological dataset - Data imputation

Missing data frequently reduce the applicability of clinically collected da ta in research requiring multivariate statistics. In data imputation, missi ng values are replaced by predicted values obtained from models based on au xiliary information. Our aim was to complete a clinical child neuropsycholo gical data set containing 5.2% of missing observations. This was to be used in research requiring multivariate statistics. We compared four data imput ation methods by artificially deleting some data. A real-donor imputation m ethod which preserved the parameter estimates and which predicted the obser ved values with acceptable accuracy was used to complete the data set. In a ddressing the lack of studies with regard to treatment of missing data in n europsychological data sets, this study presents information on the outcome s of applying data imputation methods to such data. The imputation modeling described can be applied to a variety of clinical neuropsychological data sets.