Identifying and Controlling Coding Errors in Information Systems

Abstract

Decision scientists and decision makers are often faced with a situation of uncertainty in the use of their methods and techniques when they must rely on data bases and information systems for accurate data. This paper reports on a research effort designed to change the uncertainty of the human element in the preparation of reports and data accuracy into a statement of risk in computer systems utilizing fields of coded data. Human errors in these data fields are dependent on the type of character used in the coding system, i.e., alpha, alphanumeric or all numeric, but independent of other factors commonly assumed to be causal. A methodology is presented for defining the upper limit of the probability of error in a coded data field, both before and after classical edit routines are invoked. Using this methodology, data accuracy can be treated as an element of risk rather than a factor to be pondered in the application of decision science techniques.