ABSTRACT

 The existence of the problem of doubtful observations has been recognized for very ong time. It is the desire to thoroughly investigate their effects, enlighten the experimenters on how to examine their existence and then introduce some degree of objectivity into their treatments that prompted the selection of the title. Apart from these. suggestions were offered as to their control. Two sets of data were employed for the study. They include: The quarterly sales of ten coca- cola products obtained from Port Harcourt Depot (1997-2006) and The examination results of forty final students in fives courses obtained form the Department of Mathematics, University of Uyo. The methods of box plot, z-score and times series were employed for the detection of outliers. The findings have revealed that the sales data, the Box- plot detected three lower Outliers but, no upper Outlier. A time plot of the data indicated two different breaks in the pattern of the observations and further confirmed the results of the BOx- plot. For the examination results, the z- score designated as extreme outliers the 'A' and 'B' grades (upper Outliers pair) in the first course and the 'F" and 'E' grades (lower Outlier- pair) in the fifth course. Two nonextreme Outliers ('E' and 'B'- grades) were detected in the second course and another two ('D' and 'C'-grades) in the fourth course. The detection was based on the discrepancy and expected frequencies. No outlier was found .n the third course - a clear justification of the grades.