The Effect of Different Statistical Analysis Method According To Interval and Ordinal Data Nature on the Psychometric Properties of Likert Scale - A Simulation Study

Abstract

     The current study aimed to reveal the effect of different statistical analysis method on the psychometric properties of Likert-type scales, the data was generated by Monte Carlo simulation that allows controlling the properties of the generated data to study its direct impact on the results. According to the mathematical ordinal nature with unequal interspaces of the Likert-type responses, it must be dealt statistically using non-parametric statistical methods. Nevertheless, researchers are accustomed to dealing with Likert-type scales as interval scales with equal distances between their response alternatives, and in an attempt to resolve this conflict between researchers, the idea of ​​this study arise, which concluded that the psychometric properties of Likert-type scales were affected when dealing with them statistically in a way that contradicts the nature of their ordinal data. While the performance of the alpha ordinal coefficient was much better and gave a better picture of the stability of the scale without bias.

     The structural validity of the data model was calculated using the confirmatory factor analysis method. Poly-Choric sections that fit ordinal data as an alternative to the basic matrix of the method, which depends on Pearson's correlation coefficients, and the study recommended avoiding dealing with Likert-type scales on the basis that they are continuous interval data as it affect the psychometric properties of structural validity and scale reliability as a result of this interaction.