ABSTRACT:In social sciences, usability of findings obtained by administration of a scale, accuracy and coherence of decisions made are closely related to psychometric properties of a scale. This study aims to investigate whether psychometric properties and attitudes of respondents differ on the number of response category, verbal or numerical rating and item distribution. To this end, related properties of a scale were changed and different versions of the same scale were used. According to the results of the study, attitudes of respondents did not significantly differ on the original, positive and negative forms of the scale, but significantly differed on the verbal and numerical forms of the scale. Gathering all the items, which are either positive or negative, under the same scale did not cause any response setting. The Cronbach alpha reliability coefficients of each version of the same scale were compared and it was found out that the increase in the number of response category caused increase in reliability of the scale. But, this increase was not statistically significant.Keywords: Rating differences, Likert-type scales, item order, number of response categories, response setting
Extended AbstractPurpose and Significance: In social sciences, usability of findings obtained by administration of a scale, accuracy and coherence of decisions made are closely related to psychometric properties of a scale. In this study, the factors that affect the reliability and validity of Likert scales were investigated. These factors can be dealt with under four headings: 1) Response set (number of categories, the numerical or verbal expression of ratings, extremity of response, etc.), 2) Carelessness, 3) Social desirability, 4) The wording of items (trueness, clearness, simplicity and clarity, moderate expressions, positive-negative balance, the layout of items, common stem, etc.) (Oskamp, 1977;Tezbaşaran, 1997;Tavşancıl, 2010). This study aims to investigate whether psychometric properties and attitudes of respondents differ on the number of response category, verbal or numerical rating and item distribution. To this end, related properties of a scale were changed and different versions of the same scale were used.