Muddles in Pentatonic Likert-type scale: Accuracy Cost in Psychometric Measurements for Small Enterprise Development

Abstract

Likert-type scale is ordinal, hence not compatible with parametric techniques. Disregard of this fact causes flawed research outputs. Enterprises get themselves in precarious situations as ultimate consumers flawed outputs. This paper is motivated by the dearth desire by entrepreneurs to make accurate and valid decisions harvested from a dependable measurement scale. Identifying the pitfalls of Likert-type scale and remedies to address the weaknesses, form the objectives of the study. The study is anchored on the Classical Test and Generation theories. Reviewing literature and from own personal experiences in assessing students’ thesis at university level in Kenya found traditional pentatonic Likert-type scale highly favored by most young researchers in enterprise development. The researchers treated the Likert scale outputs as interval data. Consequently most of them got wrong inferential techniques and findings. This study suggests transformation of ordinal data into binary data, interval or ratio before going into parametric analysis. Secondly, increase the number of points on the Likert scale, preferably to seven (7) to enhance reliability, validity, discriminating power and respondent preferences. Thirdly, adopt newest models of Likert type scale, that is; novel fuzzy Likert scale, phrase completion scale and two-stages Likert scale for measuring direction and intensity dimensions seperately. Finally, Likert type scale could be improved by Rasch analysis, too. The findings and suggestions of the study are relevant for researchers in both academic, clinical and enterprise development for attainment of the Kenya Vision 2030 .