Exploration of Likert scale in terms of continuous variable with parametric statistical methods

The Likert scale is an ordinal variable that measures the intensity of responses from research participants. It is widely used not only in social sciences, such as sociology and psychology, but also in survey-based research fields, such as nursing and public health. Among the approaches for handling the Likert-scale data, treating it as a continuous variable has been commonly used because it facilitates the application of parametric statistical methods and interpretation of results. In addition, from a perspective of statistical principle, this type of approach has been widely discussed and considered unproblematic. However, studies exploring the characteristics of the Likert scale in the approach with simulations are relatively rare. Thus, this study aimed to confirm the validity of the approach with simulation that compared the statistical characteristics of the Likert scale variable with those of variables from an assumed continuous latent distribution. In the Monte Carlo simulation study, the rectified normal distribution was specifically assumed for the continuous latent distribution. Then, type 1 error, statistical power, and correlation were compared across various situations involving both single and multiple variables. From the simulation study, we found that a Likert scale with five or more points exhibited statistical characteristics comparable to those of the variable derived from the latent continuous distribution. Based on the results, we confirmed that the Likert scale can be used as a continuous variable in various parametric statistical methods. The corresponding suggested guidelines are expected to assist researchers in research design, data analysis, and results interpretation.