COMPARISON OF PARAMETRIC AND NONPARAMETRIC TEST RESULTS: ASIAN DEMOGRAPHICS AS DATABASE

Several studies used both parametric and nonparametric tests in the statistical analysis of data. A number of findings revealed varied results when both treatments were applied. In most cases, differences were due to wrong assumptions of normality of data distribution, the insufficient sample size, the inappropriate type of variables employed, etc. With these disparities, the researcher investigated the outcomes of the parametric and nonparametric tests if their respective data requirements are met and no assumptions are violated. To show fairness in handling data and procedure, the researcher used the published data on South-East Asian demographics. Data were subjected to normality test using the normal probability plot and Kolmogorov-Smirnov test. Other data requirements were taken into account before the parametric and nonparametric statistics were applied. The parametric tests employed in this research are: (1) t-test for two correlated samples; (2) t-test for two uncorrelated samples; (3) analysis of variance or F-test for three samples; and (4)Pearson r for two correlated variables; with their corresponding nonparametric equivalents of Wilcoxon Τ, Mann-Whitney U, Kruskal Wallis H, and Spearman rho. Though findings in this study revealed consistent parametric and nonparametric test results upon comparing the four sets of samples, it is essential, however, to note the observations of previous studies that the use of nonparametric tests comes at a cost in cases where a parametric test would be appropriate. The fewer assumptions in nonparametric tests make them less powerful than their parametric equivalents. Parametric test should be applied if normality is established because it gives a better chance of finding significances when they exist. Otherwise, a nonparametric test is a reasonable option.