A Likert Scale: Definition, Examples, and Analysis

A Likert scale is a psychological measure for assessing attitudes, values, and opinions. It works by having a person fill out a questionnaire in which they must rate how much they agree or disagree with a series of assertions. Rensis Likert established the Likert scale in 1932, and it is named after him. Likert scales are the most popular Type of scale used in survey research.

Examples of Likert Scales

Likert scales are essential because they provide multiple pre-written answer possibilities that apply to a wide range of circumstances, from customer satisfaction to public opinion research, in addition to the granularity they bring to survey research. For example, the ‘disagree to agree’ Likert scale (shown below) can be used to ask respondents to score their level of agreement with statements on brand affinity, political beliefs, and other topics. The following are some of the most used examples of 5 point Likert scale:

Agree to Disagree Likert Scale

•       Strongly Disagree
•       Disagree
•       Neither agree nor disagree
•       Agree
•       Strongly Agree

Satisfaction Likert Scale

•       Very dissatisfied
•       Somewhat dissatisfied
•       Neither dissatisfied nor satisfied
•       Somewhat satisfied
•       Very satisfied

Likelihood Likert Scale

•       Very unlikely
•       Somewhat unlikely
•       Neither likely nor unlikely
•       Somewhat likely
•       Very likely

Good to bad Likert Scale

•       Very poor
•       Poor
•       Average
•       Good
•       Excellent

Frequency Likert Scale

•       Never
•       Rarely
•       Sometimes
•       Often
•       Always

 DATA AND ANALYSIS FROM A LIKERT SCALE

 Researchers frequently use surveys to assess and evaluate the quality of products and services. A standard classification format for studies is the Likert scale. Respondents rate the quality of a product/service on a scale of high to low or better to worse, using two, four, five, or seven levels.

For further examination, researchers and auditors usually arrange acquired data into a hierarchy of four basic measurement levels – nominal, ordinal, interval, and ratio measurement levels:

•       Nominal data: is data in which the answers are grouped into variables but do not have to have quantitative data or order.
•       Ordinal data: is information that can be sorted or classified but cannot be measured in terms of distance.
•       Interval data is a type of aggregation data in which orders and distances can be measured.
•       Ratio data: is a type of interval data that is comparable to interval data. The sole distinction is that each data set has an equal and definitive ratio, whereas absolute “zero” is viewed as a point of origin.

The examination of the nominal, interval, and ratio data is often straightforward and transparent. Ordinal data is used to evaluate data, notably in surveys with Likert or other scales. This isn’t a new issue. In survey analysis in numerous applied fields, the usefulness of treating ordinal data as interval data is still controversial. The following are some important considerations:

Statistical tests include: Because parametric statistical tests are more powerful than nonparametric alternatives, researchers sometimes consider ordinal data interval data. Furthermore, parametric test inferences are simple to comprehend and provide more information than nonparametric tests.

Concentration on Likert scales: However, treating ordinal data as interval data without first considering the data set’s values and the study’s aims can lead to misinformation and misrepresentation of survey results. Researchers prefer to treat ordinal data as interval data and concentrate on Likert scales when analyzing scalar data.

For data inspection, use the median or range: When the data is on ordinal scales, a universal guideline implies that the mean and standard deviation, like any parametric analysis based on the normal distribution, are meaningless parameters for detailed statistics. The nonparametric test is performed on data using the appropriate median or range.

Best techniques for assessing Likert scale results

Because Likert data is discrete, ordinal, and limited in scope, there has long been a debate about the best logical way to evaluate it. Parametric and nonparametric tests are the first two options. The following are general descriptions of the benefits and drawbacks of each style of analysis:

Parametric tests presume that division is done in a consistent and orderly manner.

Nonparametric tests do not presume that division occurs in a regular or unbroken pattern. However, there are fears that the ability to notice a difference when one exists would be harmed.

 Which is the most suitable option? When it comes to analyzing data from a survey that employs Likert Scale questions, this is a fundamental decision that a researcher must make.

•       Many research has attempted to answer this topic over the years. However, they tend to examine a small number of alternative distributions for Likert data, causing the conclusions to be less generalizable. Simulation studies can now properly analyze a wide range of distributions because of advances in computational capacity.

•       The researchers discovered 14 different distributions that are typical of actual Likert data. To test all conceivable combinations of the 14 distributions, the computer software extracted self-sufficient pairs of samples.
•       For each of the 98 distribution combinations, 10,000 random samples were created. To compare the efficacy of each test, the samples’ pairings are evaluated using both the two-sample t-test and the Mann-Whitney test. The study also looked at various sample sizes.

•       The results reveal that the Type I error rates (false positives) for all pairs of distributions are incredibly close to the target numbers, indicating that if an organization employs any of the studies and the results are statistically significant, it does not need to be concerned about false positives.

•       The results also reveal that the difference in power between the two tests is minimal for most combinations of distributions. If a discrepancy exists at the population level, each analysis has an equal chance of detecting it.

•       There are several pairs of special distributions where the two tests have different power. If an organization does two tests on the same data and the results aren’t the same (one is significant, the other isn’t), the power difference affects just a tiny percentage of cases.
•       The decision between the two analyzes is, in general, a loop. When a company has to compare two sets of five-point Likert data, the analysis technique usually doesn’t matter.

•       Both parametric and nonparametric tests consistently provide the same level of protection against false negatives and false positives. These tendencies hold for groups of 10, 30, and 200 people.

Rather than assuming that an experience was satisfactory or that a respondent will positively react to a statement, researchers should ensure that the report or statements being evaluated are clear, the answer options are distinct from each other (as demonstrated in the above pre-written choices), and the scale encompasses all possible options.

 

About the author:

Lori Gillen is a Blogger/Content Creator who is specialized in the field of Digital Marketing & Data Analysis with 5 years of experience. Currently working at PPCexpo as a Senior Content Creator.