The foregoing has been a brief description, using a hypothetical example, of the process followed by the statistician in making an estimate from a statistical sample, in specifying the accuracy of the estimates in terms of reliability and precision, and in designing the sample. As a brief description, the discussion has been restricted to the case of simple unrestricted random sampling and is necessarily simplified. The full complexities of statistical theory have been avoided and the many practical problems which are faced in instituting a statistical sampling plan have not been discussed. Despite omissions, it is hoped that the accountant has gained some insight into statistical sampling procedures and the language of the statistician. Particularly, it is hoped that the accountant has gained understanding of the concepts of precision and reliability which the statistician uses in defining accuracy. Known accuracy--known precision and reliability--is one of the major advantages of statistical sampling. When the accountant uses statistical sampling, he can benefit from this advantage of determinable accuracy by specifying the levels of precision and reliability he desires for the estimate being made. However, if these basic concepts of statistical sampling are not understood, the advantages of statistical sampling may not be realized.

The accounting systems of business have traditionally provided much of the financial data and much of the analysis of that data applicable to decision-making purposes. Today, however, there are symptoms of management dissatisfaction with current accounting systems. Accounting for decision-making involves a particular way of viewing the decision-making, or managing process in business. Choice of a best alternative requires criteria against which to judge various possibilities. Choice requires information concerning the various alternatives, information cast in a form consistent with the criteria. It has been assumed that the decisions to he made by an organization must first he specified before an intelligent design of the information flow is possible. Determination of the decisions to be made comes properly before specification of the data which should be supplied. An organization which does not recognize that it has a poor system for routine inventory decisions will suffer reduced profits. Accounting has an obligation to take a significant part in the development of new quantitative information systems. Accounting must divert itself from its preoccupation of the past with fiduciary and stewardship responsibilities.

The article focuses on a mixed-sequential estimating procedure with application to audit tests in accounting. One of the purposes in external auditing is to develop a valid opinion on the state of control of a business system. To arrive at such an opinion involves detailed system tests, for example, testing invoices against purchase orders with respect to prices and quantities. Because of time and expense, it is desirable to minimize the extent of these tests, consistent with the quality and standards of assurance desired. In this commonplace audit situation, the use of various statistical sampling techniques has been long advocated. In practice, however, few accountants have found existing sampling techniques to be entirely satisfactory in evaluating the state of system control. The purpose of a pure acceptance sampling technique is essentially to make a decision. As applied to a particular control area, acceptance sampling will yield a "yes" or "no" decision as to whether or not the control area meets a pre-specified level of acceptability. In any particular control area, the auditor, however, is rarely interested in making a yes or no decision on acceptability.

Reviews the book "Management Controls: New Directions in Basic Research," edited by Charles P. Bonini, Robert K. Jaedicke and Harvey M. Wagner.