In dealing with certain types of psychological measures, particularly those involving perception and attention, one is sometimes confronted with excessive values which are probably due to lapse of one type or another. Occasionally these extreme values may be numerous enough to spuriously influence the results. This is particularly true when the mean value of a series of scores is taken as a basis for comparing performance of two or more individuals or groups under different experimental conditions. Several rule-of-thumb methods have been used to secure a valid index for comparison of such measures. Among them are: 1. Use of the median which is not influenced by long responses. 2. Use of the mode as being a typical score. 3. To arbitrarily throw out the unduly long measures as being non-representative or atypical. 4. To average the long measures in with the others assuming that they are representative or typical. 5. To assign an average score to extreme measures. While the median gives little weight to extreme deviations, it may fail entirely to represent the type. The mode is perhaps in theory the most typical score. However, if an array is very irregular, there is, in strictness, no mode or type at all, or at least the indicated mode has little significance.
Proceedings of the Iowa Academy of Science
©1955 Iowa Academy of Science, Inc.
Suhr, Virtus W.
"A Comparison of Methods for Handling Extreme Values in Experimental Procedure,"
Proceedings of the Iowa Academy of Science, 62(1), 473-478.
Available at: https://scholarworks.uni.edu/pias/vol62/iss1/58