In this article some of the causes of matrix singularity are discussed. A relatively simple multivariate procedure is developed to study the variable structure of singular data. This procedure, referred to as data raking, identifies the variable(s) causing singularity. Once identified the variable(s) may be eliminated from the data resulting in the desired non-singular data set that is usually assumed by most multivariate procedures. The logic of the procedure is clarified with illustrative examples using both singular and non-singular data.