A key consideration in designing surveys or analyzing research data is the manner in which the responses that one collects should be classified. In statistics, the measurement of the data directly determines legitimate methods of analysis and possible interpretation of the results. This category is commonly known as the level of measurement and the two most commonly discussed are the Ordinal vs Interval data and the interval.
The Likert Garrett scale, now commonly applied in any survey design to capture the feelings of preference, satisfaction and opinions, takes the center-stage of its impact. According to some researchers, Likert data is ordinal only, but in order to make it interval, they should be measured under certain conditions. It is important since it helps to decide whether you have to apply rather basic ranking approaches, or more sophisticated statistic analysis. This blog will discuss the differences between ordinal vs interval scales, how to classify the Likert scale, and what best practices around treating Likert responses in research are.
Types of Measurement Scales in Statistics
The statistical information is categorised into various levels of measurement that define what type of analysis can be executed.
1: Nominal Scale
Nominal scale is the simple type of measuring scale. It stores data in categories without ranking or orderly listings An example of nominal variable is gender, blood type, or hair color as they do not imply that one type of data is either superior or inferior. Nominal scales could be applied to classification and not ranking or other calculations as one could only count numbers of classes.
2: Ordinal Scale
The ordinal scale provides order in terms of ranking categories but the distances between the categories are not always equal. An example is the satisfaction ratings as poor, fair, good and excellent because it can be noticed that as one goes up the scale of satisfaction (ie. The relation between poor and fair; Fair and good; Good and excellent), the level of satisfaction may not be the same. This means ordinal vs interval data can be applied when showing trends or preferences, but, it is of limited applicability when it comes to accurate measurement.
3: Interval Scale
The interval scale goes one step ahead and attempts to make sure that not only the categories are arranged and ordered but also that there is spacing between the values. An everyday example is temperature in Celsius or Fahrenheit in which the difference between 10degC and 20degC is also the same as the difference between 20degC and 30degC. Interval scales however, do not have an absolute zero point and as such, ratios cannot be made. Values are not comparable in terms of doubling, halving, etc. and one cannot add or subtract.
4: Ratio Scale
The most advanced measurement is the ratio scale. With all the properties of interval data of ordering and even spacing, it has a meaningful absolute zero. Examples are the weight, height and age. Ratio data is also supportive of all functions of mathematics including ratios due to its true zero point. As an example, an individual, aged 40, is definitely two years old than another individual, aged 20, and as such, ratio data is the most flexible in performing statistical analysis.
These kinds of measurements scales in statistics are the basis of a good research design and analysis.
Ordinal Scale Examples
The ordinal scale is quite common in surveys and studies in which order may be important but not measurable differences.
Characteristics of Ordinal Data
- Categories have a significant order
- The distance between categories are not uniform
- The medians and the modes are true, whereas the means can be deceptive.
Common Examples
- Level of education (high school, bachelor, Master, doctor).
- Satisfaction reports (unsatisfied, neutral, satisfied).
- The class ranks (down, middle, up).
Ordinal scales are great at displaying trends but with their statistical drawbacks, they cannot be employed in complex calculations.
Interval Scale Definition
The interval scale is a more advanced scale of measurement due to the consistency and meaningfulness of the differences between values.
Characteristics of Interval Data
- The values are equidistant
- There is no zero point, (zero is not nothing).
- Permits addition and subtraction as well as ratios.
Common Examples
- Temperature in Celsius or Fahrenheit.
- Calendar years.
- Minimally standardized performance test results
Incidences data allows more complex statistical analyses, such as correlation and regression, because incidences data assumes that there is equal spacing between categories.
Likert Scale Data Type
One of the most widely used in survey designing tools is the Likert scale. The respondents are required to rate their level of agreement or satisfaction on a scaled number (e.g. 1 = strongly disagree to 5 = strongly agree).
Characteristics of Likert Data
- Gives Categorical answers
- Includes a Midpoint (Auto Setting).
- Is typically in either a 5-point or 7-point form.
The controversy is due to the fact that a Likert scale possesses ordinal characteristics (order of the categories) but often is applied in an interval manner in real practice.
Ordinal vs Interval Scale in Research
Likert as Ordinal
Likert responses can be treated as ordinal and are analyzed as a series of ranks. Agree is larger than neutral and neutral is larger than disagree, but the difference between categories is not known. The methodology is effective in medians, modes and non-parametric tests of statistics.
Likert as Interval
When considered an interval variable, Likert responses are supposed to have spaces that are equal between categories. With it, researchers are able to calculate means and standard deviations and apply parametric tests like t-tests, ANOVA and regression.
- This discussion remains open, but the consensus of researchers is that although Likert items are ordinal, Likert scales that have been summed or averaged can usually be treated as interval.
Statistical Analysis of Likert Data
Depending upon your classification of Likert data to be either ordinal vs interval, the selection of the right analysis is made.
When Treated as Ordinal
- Use non-parametric tests (Mann-Whitney U, Kruskal-Wallis).
- Report medians and frequency distributions.
When Treated as Interval
- Use parametric tests (t-tests, ANOVA, regression).
- Report means and standard deviations.
Both of these options are legitimate, but the decision depending on the purpose of the research, the quality of the data and assumptions should be taken into account.
Quantitative vs Qualitative Data in Likert Scales
A Likert scales tests quantitative facts (numbers on a scale) but tests qualitative point of view, opinion, or attitude.
- Quantitative: Uses numerical coding, which is calculatable.
- Qualitative: The answers are based on subjective judgments.
This duality of Likert scales contributes to the high flexibility of the scales and yet controversial in statistics arguments.
Treating Likert Scale as Interval Data
Popularity has led many researchers to apply Likert scale data as interval, pragmatically:
- Big samples minimize the effect of the measurement error.
- Social sciences normally use the averaging of responses, with the assumption of equal displacement.
- It enables deeper insights with powerful statistical analysis
Nevertheless, this supposition should be explained. When data are misclassified, false conclusions might be developed.
Scale of Measurement in Survey Design
An important aspect of survey design is the choice of the scale of measurement that should be reliable and valid.
- Ordinal scales should be used when an order is important, but precision is not.
- Use interval scales where it is necessary to have an equal spacing between values
- Be careful when coding Likert scales- be mindful of what you intend to analyse
Well-thought out survey design enhances the accuracy of data, satisfaction of the respondents, and quality of research.
Conclusion
There is no exact answer to the question of whether a Likert scale is ordinal vs interval. Likert items are technically ordinal but they are often treated as interval data so that richer analysis can be performed. Learning the measurement scales in statistics, including examples of ordinal vs interval scales and the definition of interval scales, will assist you in making better surveys and picking the correct analysis techniques.
When your interest is plain ranking and trend detection, then Likert scales can be thought of as ordinal data. As part of a large sampled analysis, interval data may be at least reasonable. The trick, ultimately, is to make your level of measurement appropriate to the objectives of your research. On making this move, you will be enhancing the validity of your surveys and deriving actionable data that could be used in decision-making.
