Summarizability in OLAP and Statistical Data Bases

Summarizability of OLAP and Statistical Databases is an a extremely important property because violating this condition can lead to erroneous conclusions and decisions. In this paper we explore the conditions for summarizability. We introduce a framework for specifying precisely the context in which statistical objects are defined. We use a three step process to define normalized statistical objects. Using this framework, we identify three necessary conditions for summarizability. We provided specific tests for each of the conditions that can be verified either from semantic knowledge, or by checking the statistical database itself. We also provide the reasoning for our belief that these three summarizability conditions are sufficient as well.