It is necessary to check an outlier or outliers on data processing for estimating population mean from analytical observations. For this purpose, there are eleven kinds of checking procedures prepared as ISO and JIS standard by our survey. In this paper, it was made clear as follows: 1) The eleven checking procedures specified in ISO standards and JIS can be systematized as such a set of P-series and its variation which consist of the simplest procedure and its supplementary procedure to make up with its weakness. 2) Among the procedures, under the model where one outlier exists in a set of n observations initially collected, the one which is most unlikely to be effected by an outlier is a procedure P(0;n) without adding any observations and the procedure P(0;n) could not be generally imploved. 3) The procedure P(0;n) is also the one less effected by an outlier when the probability for the occurrence of an outlier is large, in such cases as an outlier occurs with a certain probability a) by a set. b) by an observation.
View full abstract