Publication Date: 6/1/74     Pages: 2     Date Entered: 2/22/84     Title: RECOMMENDED PRACTICE FOR DEALING WITH OUTLYING OBSERVATIONS     June 1974     U.S. ATOMIC ENERGY COMMISSION     REGULATORY GUIDE     DIRECTORATE OF REGULATORY STANDARDS     REGULATORY GUIDE 5.36     RECOMMENDED PRACTICE FOR DEALING WITH OUTLYING OBSERVATIONS A. INTRODUCTION     Section 70.51, "Material Balance, Inventory, and Records     Requirements," of 10 CFR Part 70, "Special Nuclear Material," requires     certain licensees authorized to possess special nuclear material to     establish and maintain sufficient written material control and     accounting procedures to enable the licensee to account for the special     nuclear material in his possession. Section 70.22, "Contents of     Applications," requires applicants for certain AEC licenses for special     nuclear material to submit to the Commission as part of the application     a full description of such procedures, including recording, reporting,     physical inventory, and measurement procedures for special nuclear     material. Essential to satisfactory material control and accounting     procedures are measurement and statistical control. This guide     identifies methods and procedures acceptable to the Regulatory staff for     dealing with the problem of outlying observations in samples and for     testing their statistical significance. B. DISCUSSION     Committee E-11 on Statistical Methods of the American Society for     Testing and Materials (ASTM) has revised a standard that deals with the     problem of outlying observations in samples and how to test their     statistical significance. An outlying observation, or "outlier," is one     that appears to deviate markedly from other members of the sample in     which it occurs. In this connection, the following two alternatives are     of interest: 1. An outlying observation may be merely an extreme manifestation of     the random variability inherent in the data. If this is the case, the     value should be retained and processed in the same manner as the other     observations in the sample. 2. On the other hand, an outlying observation may be the result of     gross deviation from prescribed experimental procedure or an error in     calculating or recording the numerical value. In such cases, it may be     desirable to institute an investigation to ascertain the reason for the     aberrant value. The observation may even eventually be rejected as a     result of the investigation, though not necessarily so. At any rate, in     subsequent data analysis, the outlier or outliers will be recognized as     probably being from a different process than that of the sample values.     The committee's purpose in developing this revised standard is to     provide statistical rules that will lead the experimenter almost     unerringly to look for causes of outliers when they really exist, and     hence to decide whether alternative 1 above is the more plausible     hypothesis to accept, as compared with alternative 2, in order that the     most appropriate action in further data analysis may be taken. The     procedures covered therein apply primarily to the simplest kind of     experimental data, that is, replicate measurement of some property of a     given material or observations in a supposedly single random sample.     Nevertheless, the tests suggested do cover a wide enough range of cases     in practice to have broad utility.     Topics addressed in this standard are:     a. Basis of statistical criteria for outliers     b. A single observation in a sample     c. The least and the greatest observation in a sample     d. The two largest or the two smallest observations as probable     outliers     e. Rejection of several outliers     f. Recommended criterion using independent standard deviation     g. Recommended criteria for known standard deviation     The revised standard was approved by ASTM Committee E-11 in     February 1974 and will be published in August 1974 with the designation     E178-74.(1)C. REGULATORY POSITION     The recommended practices dealing with the problem of outlying     observations in samples and the methods for testing their statistical     significance contained in ASTM Standard E178-74,(1) "Recommended     Practice for Dealing with Outlying Observations," are generally     acceptable and provide an adequate basis for determining if an outlying     observation is an aberrant value or an extreme manifestation of the     random variability inherent in the data, subject to the following     qualifications:     ----------     (1) Copies may be obtained from the American Society for Testing     and Materials, 1916 Race Street, Philadelphia, Pennsylvania 19103.     ---------- 1. In regard to the statistical criteria discussed in Section 3.1 of     the ASTM Standard, the level of significance should not exceed 0.5     percent. 2. Statistical outlier tests should be applied to basic measurement     control processes, but not directly to the rejection of observed MUF     values. 3. Caution should be applied to avoid unwarranted rejection of     suspect observations. Discarding an overly large number of false     outliers could introduce bias rather than eliminate it. 4. In regard to Section 3.2, the assumption of normality should be     assessed using Regulatory Guide 5.22, "Assessment of the Assumption of     Normality (Employing Individual Observed Values)," April 1974. 5. All discarded outliers should be recorded as part of the records     maintained for special nuclear material control purposes.     32