Despite the existence of numerous meanings, outliers are understood as data points which are far outside the standard for a population or a variable. The outlier could be an observation or a reading that is too different from other readings to the extent that it raises suspicion in regards to the approach used in data collection. Due to the relative nearness of outliers to the distribution center, they could lead to an unreasonably powerful effect on parameter approximations. When carrying out statistical analyses, outliers can cause deadly effects lowering statistical power tests, exaggerating error variance, decrease normality, and extremely influence important estimates (Jason " Amy, 2004).
Outliers mainly arise due to data errors, intentional misreporting, sampling error, standardization failure, and defective distributional assumptions. Human error can be a major cause of outliers during data entry, recording, and collection. Additionally, some participants in surveys are likely to intentionally provide incorrect data, especially when dealing with sensitive data and there is the need for self-presentation and social desirability (Jason " Amy, 2004). For instance, motivated over-reporting is likely to occur in teenagers when the subject is socially desirable, including sexual experience, church attendance, study time, and grades among others. Outliers could also arise when some members of a sample are accidentally from a diverse population than the rest of the sample.
The extreme values on a single variable are known as univariate outliers. For instance in a psychological study where there are five survey questions, one would conduct five distinct univariate outlier analyses with each study group, gender or any other difference among the participants. When carrying out research or a study that involves more than one condition, like manipulating sadness and happiness, the researcher should conduct a univariate analysis on the five questions within both groups.
In a roughly normal set of data, the percentage of scores within one standard deviation of the mean account for “roughly 68% of the set, about 95% of the set within the two standard deviations, and 99.7% of the set within the three standard deviations” (Westfall " Henning, 2013, p. 243). According to Westfall and Henning (2013) the 68-95-99.7 rule is used in statistics to help in remembering the percentages of scores in a normal data set. In social science, the 99.7% probability is treated empirically as a near certainty value.
In statistics, skewness represents the measure of asymmetry and an imbalance from the mean of data distribution. In various situations, skewed data arises pretty naturally. For instance, income is generally skewed positively since only a few wealthy individuals with huge income are likely to affect the mean, and it is unlikely to have a negative income. Most data sets form a distribution shape that has one peak and more likely to be skewed or symmetric to one side (Schinka, Velicer, " Weiner, 2013). A distribution is skewed positively or to the right when its shape has a longer right tail, and the bulk of the data is to the left and vice versa for a negatively or left distribution. For positively skewed data, the median and the mean are always greater than the mode. Also, the general rule for positively skewed data is that the mean will always be greater than the median (Schinka, Velicer, " Weiner, 2013). On the other hand, for the negatively skewed data, the mode is greater than the median and the mean. In addition, the general rule for the data skewed to the left is that the median will always be greater than the mean (Schinka, Velicer, " Weiner, 2013).
References
Jason W. Osborne, " Amy Overbay. (March 01, 2004). The power of outliers (and why researchers should ALWAYS check for them). Practical Assessment, Research " Evaluation, 9, 6, 1-8.
Schinka, J. A., Velicer, W. F., " Weiner, I. B. (2013). Research methods in psychology. Hoboken, N.J: Wiley.
Westfall, P. H., " Henning, K. S. S. (2013). Understanding advanced statistical methods. Boca Raton, FL: CRC Press.
