SOME NEW STATISTICAL CONCEPTS FOR QUALITY CONTROL R. R. Kinniston and A. N. Jarvis Environmental Monitoring and Support Laboratory Las Vegas, Nevada ABSTRACT A statistical technique is presented whereby radioisotope quality control can be substantially improved. The basis of this technique 1s an examination of the statistical distribution of the data. INTRODUCTION The rule generally used in most quality control work for setting acceptance limits is to arbitrarily accept some multiple of the standard deviation and use this along with the mean value to define acceptance or control limits. Currently there is significant discussion about what multiple of the standard deviation should be used and why one should choose any specified value. What is usually overlooked in these discussions 1s that this commonly used rule contains the inherent assumption that the data are normally distributed. The Most frequently used multiple is three standard deviations which, if the assumption of normality is correct, implies a 1/4 percent probability that a good data value will be judged to be out of control. We believe that the current discussion, which concerns what this probability value should be, is of minor consequence. What is of consequence is that the data satisfy the assumption of having a normal distribution. If this assumption is not reasonable for a data set, then a rule using a multiple of the standard deviation yleids statistically invalid results, and some type of more sophisticated statistical algorithm must be developed. 593