Outliers are data point(s) that deviate so much from the others that suspicions of having been generated through a different mechanism from the rest are aroused, or an indication of an error having been committed while compiling the data (Cressie, 2015). Special cases of outliers are fringeliers, which are data points that lie relatively closer to the to the distribution center. Since they lie in close to the three standard deviations from the mean, fringeliers have a strong influence on the parameter estimates. However, they are harder to identify, unlike the outliers and are much less noticeable. The contents of this research paper will attempt to delve into the details of how outliers can become part of data collected and also how outliers can affect the data set.
How Outliers can become part of the data collected.
There are different mechanisms through which outliers become part of the data collected and compiled. For example, they may be caused by unintended human errors at the point of data collection or entry. Besides, human error may result from misreporting of the data collected. In addition, intentional misreporting may also give rise to outliers in data. For instance, respondents may knowingly give inaccurate responses during a survey, motivated by different factors like the sensitivity of the subject matter, e.g., drugs uptake, the environmental conditions of the respondents or the nature of desirability of the variable (Cressie, 2015). Outliers occur when some respondents give misleading information to the researcher, while the others are honest.
Similarly, outliers may result from a sampling error, whereby the respondents selected were not fully representative of the whole section under survey. Consequently, the results of the survey are not reflective of the whole section under consideration, giving rise to the abnormalities when the data is compiled.
Also, outliers may occur as a result of inaccurate distributional assumptions. Long term or short term trends may affect the data collected in diverse ways. Library attendance, for example, may peak when the exam mood is near, and when a survey is conducted at this time, inaccurate data may be collected, giving rise to possible outliers (Cressie, 2015). Lastly, outliers may be a part of the data collected when there is a legitimate, random sampling of the population. The section to be put under survey may be randomly selected and still wild deviations from the mean recorded and when data is compiled, the outliers discovered.
How Outliers Affect the Data Set
The error variance is increased due to outliers, and the power of the statistical tests diluted. The effect on the mean is significant on the statistical data. For example, with a higher outlier, the overall mean increases while a lower mean is noted with a lower outlier. Also, outliers have no significant effect on the median while registering no impact at all on the mode. Despite all these, outliers affect the range by increasing the spread of data. Secondly, if the outliers are not randomly distributed, they have the effect of decreasing normality (Silverman, 2018). They may also have severe influences on estimates that may be of substantive interest.
Conclusion
It is clear that outliers the affect statistical data analysis in a significant manner. As a result, it is essential that the researcher has to devise ways of discovering them and identifying the reason that the outliers are present in the data analysis. It is of particular importance that the factors that give rise to the outliers are given attention since they could lead the researcher into the discovery of more underlying aspects of the section under survey.
References
Cressie, N. (2015). Statistics for spatial data. John Wiley " Sons.
Silverman, B. W. (2018). Density estimation for statistics and data analysis. Routledge.
