Issues in Statistical Data Analysis in Applied Linguistics: Data Normality, Transformations and Power Analysis
Lei He · US-China Foreign Language · 2011
This paper invites researchers of applied linguistics dealing with quantitative data to the issues that have been in oblivion in statistical analysis. These issues include the evaluation of data normality and data transformations, as well as power analysis with the associated estimation of sample sizes and calculation of effect sizes. Methods introduced in this paper to test data normality include calculating descriptive statistics and performing the Kolmogorov-Smirnov test. Three ways to transform non-normal data are provided: the arcsine, square root and natural log transformations. A method to reflect negatively skewed data is also included. In addition, power and its related sample size and effect size analyses are introduced in the end. The applications of these methods are illustrated by examples.