Abstract:

In statistics, it is assumed that the data suitable for parametric tests exhibit a normal distribution. In other words, before running parametric tests, it is essential to check whether the data set follows a normal distribution. Parametric tests can only be applied to data that adhere to a normal distribution. The Shapiro-Wilk and Kolmogorov-Smirnov (Lilliefors modification) tests are among the most commonly utilised methods for testing normality. However, these tests were originally developed for rational numbers. The use of these tests for Likert-type digital or face-to-face survey data has always been a topic of discussion. Even if the measurement tools consist of at least three items, the data sets are dominated by repeated values. Again, there are different opinions on the acceptable range for kurtosis and skewness values. With this study, it was determined that the Kolmogorov-Smirnov (Lilliefors modification) test is not suitable for testing the normality of Likert-type digital or face-to-face survey data. In addition, the coefficients of kurtosis and skewness within the range of (-1, +1) range are acceptable for normality. For this reason, it is recommended not to use goodness-of-fit tests such as Shapiro-Wilk or Kolmogorov-Smirnov for testing the normality of interval scale data collected through questionnaires.

Keywords: