Many real-world phenomena generate data sets with outliers i.e., extreme observations that are away from the mainstream of the data. The presence of outliers may cause invalid analysis by violating the conventional assumptions of regression models. Hence identification of outliers holds significant importance in data analysis. This study reviews various outlier labeling methods and shows the comparative detection of outliers by applying these methods on several real data sets with small to large sample sizes and low to high levels of skewness. Some graphical and formal methods of univariate outlier detection are also applied. All labeling methods detected no outlier for symmetric shape except adjusted boxplot. For slightly skewed distribution, Z-score, 3SD method, and 3IQR found resistance for both small and large sample sizes except adjusted boxplot which is resistant in large data only. In the case of mildly skewed and large sample size, the 2Median Absolute Deviation method shown up most sensitive. It is concluded that the Adjusted boxplot, Z-score, 3SD method, and Tukey's 3IQR (interquartile range) method detected fewer outliers among other competing methods. Boxplot and a formal Generalized ESD test identified outlying observations as well as most extreme observations.