Understanding Scales of Measurement in Educational Statistics

When a teacher records whether students passed or failed, ranks them by performance, or averages their test scores, they are working with data – but not all data is the same. The type of data you collect determines which statistical tools you can legitimately use to analyze it. This is where scales of measurement become essential. Developed by psychologist Stanley Smith Stevens in a landmark 1946 paper, the four scales – nominal, ordinal, interval, and ratio – form the backbone of educational statistics. Understanding them is not just academic; it directly shapes how you calculate and interpret the mean, median, and mode in any assessment context.

What are scales of measurement? 🔗

A scale of measurement describes the nature of the information contained in a variable’s values. In research, the scale used for measurement determines what kind of analysis is appropriate and what conclusions can be drawn from the data. Put simply, before you calculate any statistic – a class average, a middle score, a most common grade – you need to know what kind of data you are working with. The four scales are often described as additive: each higher scale retains the properties of the scale below it and adds something new.

The four scales and how they work in education 🔗

Nominal scale: labeling without ranking 🔗

The nominal scale is the simplest form of measurement. It is used purely for classification – grouping data into distinct categories that have no numerical relationship or natural order. In an educational setting, examples include a student’s preferred subject (Mathematics, Science, History), the type of school they attend, or their gender. The categories are mutually exclusive and carry no implication that one is greater or lesser than another.

Nominal scales simply name groups based on characteristics, with no specific order. Because of this, the only measure of central tendency that makes sense for nominal data is the mode – the category that appears most often. The mode is the only measure of central tendency that can be used for data measured on a nominal scale. Calculating a mean or median here would be meaningless, because there is no inherent numerical relationship between the labels.

Ordinal scale: ranking without equal intervals 🔗

The ordinal scale goes a step further by introducing order. Data can be ranked from highest to lowest, but the gaps between ranks are not necessarily equal or measurable. A familiar educational example is a Likert-scale survey item, where students rate their satisfaction as “strongly agree, agree, neutral, disagree, strongly disagree.” Another example is ranking students 1st through 10th by exam performance. We know that 1st is better than 2nd, but we cannot say by exactly how much.

While the order is crucial, the differences between values might not be consistent – you can rank the values, but you don’t know the relative degrees of difference between them. For ordinal data, the appropriate measures of central tendency are the median and the mode. The median is particularly valuable here because it identifies the middle point of an ordered dataset without assuming equal spacing. The median is preferred when data are measured on an ordinal scale. The mean should not be applied, because it requires equal intervals between values – a property ordinal data does not have.

Interval scale: equal intervals, no true zero 🔗

The interval scale offers considerably more precision. Not only is data ordered, but the distance between any two consecutive values is equal and meaningful. The key limitation, however, is the absence of a true zero point – zero does not signify the complete absence of the attribute being measured.

In educational assessment, IQ scores are the classic example of interval data. IQ tests are based on interval scales, where the differences between scores are meaningful, but there is no true zero point – this allows researchers to assess differences in intelligence without implying that someone with a low score has “no intelligence.” The difference between an IQ of 100 and 110 is the same as the difference between 110 and 120, which makes arithmetic operations on such data valid. Similarly, temperature in Celsius or Fahrenheit follows an interval scale: 0°C does not mean the absence of temperature.

Because of the equal intervals, the mean, median, and mode are all appropriate for interval data. With interval scales, you can calculate the mean and standard deviation for central tendency and variability measures, respectively. However, the median and interquartile range can be more appropriate for skewed distributions. One operation that is not valid with interval data is forming ratios – you cannot say that an IQ of 120 reflects “twice the intelligence” of an IQ of 60.

Ratio scale: the highest level of measurement 🔗

The ratio scale has all the properties of the interval scale with one crucial addition: a meaningful zero point that indicates the complete absence of the quantity being measured. This makes ratios between values genuinely meaningful.

In educational contexts, raw test scores out of a fixed total (e.g., marks out of 100), attendance counts, and the number of correct answers are all ratio-scale data. A student who scores zero has genuinely answered nothing correctly. A student scoring 80 has scored twice as many marks as a student scoring 40. Ratio scales provide a wealth of possibilities when it comes to statistical analysis – variables can be meaningfully added, subtracted, multiplied, and divided. All three measures of central tendency – mean, median, and mode – are applicable, and the mean is generally the preferred choice when data are normally distributed.

How the scale determines which measure of central tendency to use 🔗

The relationship between scales of measurement and measures of central tendency is not optional – it is a matter of mathematical logic. Using the wrong measure produces results that are either meaningless or actively misleading. The mode is appropriate for all variables regardless of their level of measurement; the median works only with ordinal and interval/ratio variables; and the mean can be calculated only for interval/ratio variables.

Here is how this plays out concretely in a classroom or school setting:

When to use the mode 🔗

The mode is the only valid measure of central tendency for nominal data. If you survey 200 students about their preferred learning style – visual, auditory, or kinesthetic – the mode tells you which style is most common. There is no meaningful “average” learning style, and there is no middle value to identify. The mode can be used for any level of measurement, but it is most meaningful for nominal and ordinal levels. At higher scales, the mode can still be calculated, but it becomes less informative as data becomes more continuous.

When to use the median 🔗

The median is the measure of choice for ordinal data and is also preferred for interval or ratio data when a distribution is skewed or contains outliers. In a school context, if you are looking at student rankings in a class of 30, the median rank gives you a reliable midpoint without being distorted by extreme values at either end. The median would be a better measure of central tendency in situations where the mean is being skewed by extreme values. For example, if most students in a class score between 50 and 70 out of 100, but two students score 99, the mean will be pulled upward and misrepresent the typical performance. The median will not.

When to use the mean 🔗

The mean is the most widely used measure and is appropriate only when data is on an interval or ratio scale with a roughly symmetrical distribution. It uses every data point in its calculation, making it sensitive and precise. The mean is usually the best measure of central tendency to use when your data distribution is continuous and symmetrical. In practice, this means calculating the class average on a standardized test, or finding the mean attendance across a semester – both valid uses of the mean with ratio-scale data.

It is worth noting a common point of debate in educational measurement: Likert-scale data. Survey responses rated 1 to 5 are technically ordinal, meaning the mean should not be applied. However, some argue that so long as the unknown interval difference between ordinal scale ranks is not too variable, interval scale statistics such as means can meaningfully be used on ordinal scale variables. This is a widely practised convention in educational research, though it should be approached with awareness of the underlying limitation.

Practical implications for educational assessment 🔗

Choosing the right scale – and the right measure of central tendency – is not a formality. It shapes the validity of the conclusions drawn from assessment data. Consider a school administrator who wants to report on student performance in a term. If the data comes from ranked grades (A, B, C, D, F), this is ordinal data, and reporting the “average grade” as a mean is technically inappropriate. Reporting the median grade or the most common grade (mode) would be more accurate.

Conversely, if the data comes from raw percentage scores (ratio scale), computing the mean is entirely valid and gives the most complete picture of overall performance. Similarly, when designing surveys to evaluate teaching quality, knowing whether your response options produce nominal, ordinal, or interval data will determine how you analyze and report the results. Each scale fulfills a function of the scale preceding it – nominal scales simply name; ordinal scales name and order; interval scales name, order, and establish proportionate intervals between variables.

A quick reference summary 🔗

To consolidate everything above, here is how the four scales map to the appropriate measures of central tendency in educational assessment contexts:

Nominal (e.g., subject preference, school type) → Mode only. Ordinal (e.g., Likert scale ratings, class rankings) → Median and Mode. Interval (e.g., IQ scores, standardized test scores without a true zero) → Mean, Median, and Mode; ratios not valid. Ratio (e.g., raw marks out of 100, number of correct answers) → Mean, Median, and Mode; all mathematical operations valid.

The higher you move up the scale, the more statistical power you gain. But each scale also imposes real boundaries on what can meaningfully be said about the data – and staying within those boundaries is what makes educational assessment credible and trustworthy.

What do you think? When you look at the assessments used in your school or institution, can you identify which scale of measurement each one relies on – and whether the reported statistics actually match that scale? And if Likert-scale surveys are widely used to evaluate teaching quality, does treating that ordinal data as interval-level for calculating means feel like a reasonable compromise or a methodological shortcut worth reconsidering?

References

1. https://en.wikipedia.org/wiki/Level_of_measurement
2. https://researcher.life/blog/article/levels-of-measurement-nominal-ordinal-interval-ratio-examples/
3. https://statisticsbyjim.com/basics/nominal-ordinal-interval-ratio-scales/
4. https://pmc.ncbi.nlm.nih.gov/articles/PMC3157145/
5. https://courses.lumenlearning.com/introstats1/chapter/when-to-use-each-measure-of-central-tendency/
6. https://docmckee.com/cj/docs-research-glossary/interval-scale-definition/
7. https://www.mymarketresearchmethods.com/types-of-data-nominal-ordinal-interval-ratio/
8. https://pressbooks.bccampus.ca/simplestats/chapter/3-7/
9. https://www.scribbr.com/statistics/central-tendency/
10. https://statistics.laerd.com/statistical-guides/measures-central-tendency-mean-mode-median.php
11. https://statistics.laerd.com/statistical-guides/measures-central-tendency-mean-mode-median-faqs.php
12. https://files.eric.ed.gov/fulltext/EJ1274528.pdf