Handling of Erroneous Values

An erroneous value can be traced to a gross deviation from the prescribed procedure for establishment of reference values. Such values may either deviate significantly from the proper reference values (outliers) or be hidden in the reference distribution. Only a strict experimental protocol, with adequate controls at each step, can eliminate the latter type of erroneous values. 

Visual inspection of a histogram is a reUable method for identification of possible outhers. It is important to keep in mind, however, that values far out in the long tail of a skewed distribution may easily be misinterpreted as outhers. If the distribution is positively skewed, inspection of a histogram displaying the logarithms of the values may aid in the visual identification of outhers. 

Some outliers may also be identified by statistical tests (see Chapter 14), but no single method is capable of detecting outhers in every situation that may occur. The number of techniques suggested or recommended is, for that reason, very large. The two main problems encountered are as follows  

Gaussianity. This is the case with Dixons range test identify the extreme value as an outlier if the difference between the two highest (or lowest) values in the distribution exceeds one third of the range of ah values. 

Horn and co-workers have published a novel method in two stages for outlier detection that seems to provide a promising solution to both of the problems mentioned above. 

---