Using a Likert Scale

What is a Likert Scale?

A Likert scale is a multi-item rating system used to gather data on people's attitudes, thoughts, and opinions on specific topics. Typically, it is used as a psychometric tool in educational and social sciences for measuring opinions/agreement of subjects. Typical Likert scales include responses consisting of: strongly agree, agree, neutral, disagree, etc... Each of these responses is assigned a numeric value: strongly agree (5), agree 4 (4), neither agree or disagree (3), disagree (2), and strongly disagree (1). By giving each of the response items a value, we are able to quantify this categorical data. The use of a Likert scales is one of the most universal ways to collect survey data. People who have used this scale can thank Rensis Likert for devising such a useful measurement tool.

TalentMap Likert Scale Image

Ordinal or Interval

While researching Likert scales, one factor that seemed to cause a variety of disagreement among statisticians was determining whether the data we are dealing with should be used as interval or ordinal data (Han Ping Fung from ResearchGate). Ordinal data is similar to categorical data, but, more specifically, this data has a ranking or an order to it. For my data, this is seen in the response options ranging from strongly disagree being the lowest to strongly agree as the highest. Others describe Likert scales as Interval, meaning the values between each interval are equally spaced. My Likert scale appears to be equally spaced with the neutral option (neither agree nor disagree) ranked as a three with each other option at an equal distance from three. For example I ranked the response agree as a four, and disagree as a two, making them both the same distance away from the neutral response. Although these items appear to be “equally spaced,” the intervals between cannot necessarily be presumed as equal. What does this mean? It means that if someone responded with disagree (2), that does not mean they are twice as negative as someone who responded with agree (4), although the numbers appear to be interval. UCLA’s Institute for Digital Research and Education explains it further here.

My Data and How I Used it

I had to do some searching and reading about Likert scales in order to use one for my research. One of the biggest challenges I encountered was determining if the data I was gathering was interval or ordinal. Researchers and bloggers across the web tend to disagree on this issue (of course, the nature of the research and the questions being asked are considerable factors). In my study, I treated the data as ordinal and compared it using the mean, median, and mode of the control and experimental group responses. While some say it is not correct to find means of ordinal data, MeasuringU states “Whether someone tells you it’s permissible to take the average of ordinal data depends on their view of measurement theory.” Other blogs I have read online state that it is very common for people to use means to compare Likert scale data because it is easy to understand and simple to compare, even though it may not show you the whole picture.

By using means, I found that my control group (Group A) had an average response of 3.75 while my experimental group (Group B) averaged 3.56. This data tells me that on average, Group A responded as being more engaged throughout the Geometry lessons without the use of augmented reality (AR), compared to Group B who used AR during each lesson of the Geometry unit. This also tells me that overall, the classes were engaged throughout the lessons as their responses averaged higher than the neutral response of neither agree nor disagree (3). The mode and median for both groups stayed the same, with the response of agree (4).

Although the averages were slightly different, the median and mode show that there was not a dramatic difference in engagement for either group. To the right, you will find two graphs comparing one of the questions from the engagement survey between Group A and Group B.

*Note: Sometimes researchers will exclude the neutral category because, as you see, it is typical for respondents to select it rather than being forced to pick a side. It is also ideal to use a 5 point scale because when you have a scale larger than 5, people do not tend to pick the more extreme amounts (high or low).

See my bar chart for the self-reflection survey HERE

Google Forms Chart Images

Tips and Things to Consider

• Point choices should cover the entire measurement continuum (typically 5 points or higher) When you use 3 or less, the subjects may be forced to select an option they don’t truly align with.

• The intervals between points should be equal (The Analysis Factor).

• Never use an individual item, always a set of several items (typically added or averaged).

• Always include verbal labels (Agree, disagree, etc…).

• The number of answer choices can be even or odd.

• Include a balanced number of positive and negative responses.

"The Case for Likert's Five-Point Scale" from Talent Map is licensed under CC BY 2.0