Four Types of Scales in Research
Explain the following using examples: Ordinal scales, Preference scales, Nominal scales, Binary scales.
There are multiple different ways, means, and methods both for the collection and analysis of quantitative data for research purposes. In this paper we will look at four of these: ordinal scales, preference scales, nominal scales, and binary scales. We will use examples for potential application in a MSc dissertation.
Nominal data is counting and placing data into categories. (Denscombe, 2010, pg 243) Considering a potential dissertation focused on analyzing the factors of motivation, feelings of meaning, and expectations of deception from the perspective of people holding local elected offices in Muskegon County, Michigan, this could include a list of their positions. The levels of government considered would be: four villages, seven cities, 16 townships, and one county. The potential positions would be: village trustee, village president, village clerk, village treasurer, city trustee, city mayor, city clerk, city treasurer, township trustee, township supervisor, township clerk, township treasurer, county commissioner, prosecuting attorney, sheriff, county clerk, county treasurer, register of deeds, drain commissioner, and county surveyor. These 20 positions could be listed and the person would select which position they hold. This would be nominal data.
Ordinal data is put into a listed order without knowing the level of difference. (de Vaus, 2002, pg 204) For instance, it could be asked how much the person considers creating change as an important factor in the meaning of them holding office. They are then give a list of numbers from one to five, with one being the lowest and five the highest. Through this we can see that something is more or less important on the scale, but we can't do a more extensive analysis because we don't know to what extent.
Binary means two, and binary questions give you two options. There are three types of binary choice formats: dichotomous questions, checklists, and paired comparisons. Dichotomous questions give only two possible answers, often yes or no, more or less. In checklists you can either check the box or not, therefore they are yes or no questions, and thus binary. Paired comparisons give two options that may be descriptions to choose from. (de Vaus, 2002, pg 104-105)
Each of these could potentially be used for the dissertation survey research. For instance, one of the factors of motivation is effort. A question can be asked to assess the perception of the necessary effort needed to achieve goals before taking office versus being in office. One such question I proposed was, "Do you think your goal will take more or less effort than you expected when you were running for office?"
I haven't previously thought about using a checklist, but certain questions could be condensed that way. As in, check all that apply, since taking office you: are more worried about local politics, less worried about local politics, more worried about state politics, less worried about state politics, more worried about national politics, less worried about national politics, more worried about humanity, less worried about humanity, have been lied to, have been lied about, etc. Instead of a checklist format some of these could be used as paired comparisons too.
A preference scale could be considered like an ordinal scale, in that you agree more or agree less. You could do it differently by having a list of items that the person then numbers in the order of their preference. In the case of meaning if we take the three primary categories of creative, experiential, and attitudinal, these could be listed with a blank. The person could then number them from one to three, in whichever order of importance that they choose.
In looking at these different types of data a major consideration is what will yield the data that is useful for the context considering many factors. A 16 point checklist can be used to make sure a questionnaire is fully prepared, a shortened version of this would include: adequate time to plan, adequate time to respond, budgeted, pilot testing, clear layout, explanation of purpose, contact address, thank yous, anonymity and confidentiality, serial numbers, clear and explicit instructions, avoid duplication, clear and unambiguous, take out non-essential questions, correct order of questions, will questions produce the right kind of data? (Denscombe, 2010, pg 171)
It's important to think ahead and consider the analysis that will be done with the data. To this end there are four factors that should be considered: "the number of variables being examined; the level of measurement of the variables; whether we want to use our data for descriptive or inferential purposes; ethical responsibilities." (de Vaus, 2002, pg 203)
In my original proposal I included a total of 28 proposed questions. Many of these were binary. For instance, "Do you think your goal is more or less important than when you ran for office? Do people lie to you more or less since taking office? Has taking the position been worth the effort and hassle?" Part of my reasoning behind this is that I want people to feel like they can move through the questions fairly quickly and easily, because they are busy and if the survey takes too much of their time and effort then my response rate will be low. These questions offer only a small amount of data for simple analysis. But, by using the three main categories of motivation, meaning, and deception with multiple questions, correlations could potentially emerge that would lead to insights and greater avenues for useful study. Therefore, simple questions can yield useful data beyond the initially obvious when viewed in a greater context. A few questions are also designed to yield greater data, asking for approximate percentages in interactions for instance.
It's useful to consider the type of data desired and what will be generated when designing research. Ordinal, preference, nominal, and binary scales can all contribute when used in the correct context and for a purposeful pursuit of data and insight.
Reference List
Denscombe, Martyn (2010) The Good Research Guide. Fourth Edition. Maidenhead, Berkshire, England: Open University Press, McGraw-Hill Education.
de Vaus, David (2002) Surveys in Social Research. 5th Edition. St Leonards, NSW, Australia: Allen and Unwin.
Find more at JeffThinks.com or JeffreyAlexanderMartin.com