Testing the equality of students’ performance using Alexander-Govern test with adaptive trimmed means

Abstract

Analyzing the equality of independent group has to be done with caution. The classical approaches such as ttest for two groups and analysis of variance (ANOVA) for more than two groups always are favorable selection by researchers. However, sometime these methods were abused by the presence of nonnormality or variance heterogeneity or both. It is known that ANOVA is restricted to the assumptions of normality and homogeneity of variance. In real life data, sometimes these requirements are hard to attain. The Alexander-Govern test with adaptive trimmed mean (AG_atm) is one approach that can be chosen as alternative to the classical tests when their assumptions are violated. In this paper, the performances of AG_atm were compared to the original AG test and ANOVA using simulated and real life data. The simulation study proved that the AG_atm performs better than the original AG test and the classical test. For real life data, student’s performance in decision analysis course, measured by final examination score was chosen. Based on the exploratory data analysis, this data found to have problem of nonnormality.