A proposed metric scale for expressing opinion / Rohana Yusoff and Roziah Mohd Janor

Behavioural scientists have always coveted to find the most perfect way to measure qualitative attributes, in a way that is almost similar to measuring quantitative variables. Unlike quantitative attributes, such as length and weight, qualitative attributes do not have physical appearance which is visible and touchable, only experienced and felt by human beings. Matheson (2008) emphasized that measurement in human sciences such as in sociology is attempting to quantify the intangible. Thus measurement of qualitative attributes demand appropriate attention and consideration.
As to date there is no unanimity among researchers in behavioural sciences on a scale that could quantify intangible variables as quantitative magnitudes. The lack of a unanimously acceptable scale is identified as the root cause that sparked the ongoing debate among behavioral scientists. There is not yet a scale that is accepted by all scholars to measure or quantify or interpret attributes such as opinion, satisfaction or agreement in terms of numbers. At the moment, it is not clear exactly how scales to evaluate questionnaire items should be designed and implemented in order to reduce the random and systematic measurement errors (Sturgis, Roberts, & Smith, 2014).
Researchers have tried to overcome the lack of scale using two methods, i) apply mathematical modelling techniques to rescale data collected using scales such as Likert scale into continuous data (Granberg-Rademacker, 2010; Harwell & Gatti, 2001; Hsu, Chang, & Hung, 2007; Wu, 2007) and ii) develop a scale using straight line of various lengths called ‘Continuous Response Scale’ or ‘Visual Analogue Scales’ (Aitken, 1969; Celia & Perry, 1986; Fernando, 2003; Lerdal, Kottorp, Gay, & Lee, 2013; Munshi, 1990; Pfennings, Cohen, & van der Ploeg, 1995; Puzziawati Ab Ghani & Abdul Aziz, 2005).
In conclusion, to obtain interval data using mathematical modelling techniques, researchers must begin collecting data using either Likert scale or other ordinal scales. Besides the complex mathematical procedures, there is still a question of how accurate does the rescaled data do represent the actual data. Even though both approaches outlined above are popular among researchers they have not yet satisfied the need for a scale that could quantify qualitative attributes; explicitly, the need for an interval scale is still unfulfilled. This is the problem that this study has identified.
This study tries to find answers to several research questions: i) What properties should a scale to measure qualitative attributes in terms of numbers have? In other words what are the properties of continuous (metric) interval scale? ii) What is the best scale layout or design that will assist quick and easy data collection, and iii) Once such scale has been developed, psychometrically will it perform better than a popular existing scale used to measure qualitative attributes? That is, will the data collected using the scale have better validity and reliability coefficients?
To answer the research questions, this study has embarked on the following objectives: i) To discover identify the properties or features of a scale that could be accepted as a metric interval scale, ii) To determine propose the best scale design or layout and then develop the metric scale, and iii) To compare psychometrically, performance of developed scale with a popular existing scale to measure qualitative attributes.