Research Concepts

Scales

Often you will construct a scale to measure subject responses. A scale measures the responses on a continuum selected by the researcher. Some scales are comparative (such as ranking choices in order) and some scales are non-comparative (such as responding on a continuum from "strongly agree" to "strongly disagree).

The continuum selected by the researcher can be either quantitative or attribute. In constructing a scale, the researcher should take into account a number factors including

• does the data have a natural variable type (quantitative or qualitative);
• will the scale design influence the reliability of the responses;
• will you need to make comparative judgements;
• how will you use the scale and what statistical tests might you run;
• how many scale divisions should there be;
• should there be an even or odd number of divisions (an odd number permits a neutral or zero choice, an even number forces a preference);
• what graphical summary should be for the scale (a graph, a vertical list, a horizontal list for example)

Some examples of comparative scales are pair-wise comparisons, rank-ordering a list, constant sum scaling (for example the subjects are given $100 and able to allocate this across a choice of purchases), and social distance scaling (measures the degree to which a person is willing to be affiliated with a group having certain traits).

Some examples of non-comparative scales are continuous rating (respondents mark their response on a line), Likert scales (respondents select from a continuum ranging from strongly agree to strongly disagree), and various types of checklists.

All scales should be verified for both reliability and validity.