The author presents a new relational approach to multivariate and multidimensional OLAP. In this approach, a multivariate aggregate view (MAV) is defined. MAV contains categorized univariate and multivariate aggregated data, which can be used to support many more advanced statistical methods not currently supported by any OLAP models. The author shows that MAV can be created, materialized, and manipulated using SQL commands. Thus, it can be implemented using commercial relational DBMS. A query rewrite algorithm is also presented to convert aggregate queries to base tables into those to MAV. Therefore, users need not to know the existence and definition of MAV in order to share materialized data. Incremental update of MAV created from single base table is also considered. The application of MAV to data mining is presented to illustrate the use of multivariate and multidimensional OLAP.