Outlier is a term used in statistics to describe the phenomenon of numerical values or data that is distant from the other data obtained. In other words: "An outlying observation or outlier is one that appears to deviate markedly from other members of the sample in which it occurs" (Grubbs, 1969, p. 1). Outliers are significant in a number of ways and require specific investigation. As Grubbs has pointed out, an outlying observation "may be merely an extreme manifestation of the random variability inherent in the data" (1969, p.1). If a researcher finds this to be the case, the values should be treated like the rest of the values of the sample, without any specific adjustments.
However, an outlying observation might also be caused by a significant deviation from the prescribed experimental procedure, or it might occur as the result of a calculation error. Accordingly, researchers have to make sure that the outliers were not caused by deviations from the prescribed experimental procedure or miscalculations. If researchers finds that in fact the outliers were caused by these mistakes, the data should be dismissed or it should be noted that this data is "faulty" and a useful analysis cannot take place (Grubbs, 1969).
In his 1969 article, Grubbs (1969) outlines several statistical methods researchers can use to test whether the outliers are indeed the result of random variability within the sample, or whether they must be the result of deviations from experimental procedures and calculation errors. In either case, outliers are to be treated with utter most care, as they can potentially deem a statistical analysis useless, if they are not treated properly. One example for outliers in health-care has been provided by a study that showed outliers in death rates in hospitals (Blumberg, 1987). Statistical analysis showed that certain hospitals had significantly higher death rates of patients. The outlier data was used to
...