What is a level of measurement?

A level of measurement describes how a variable is recorded. It tells you whether values are only labels, whether they can be placed in order, whether the gaps between them are equal, and whether zero means none of the measured amount.

This idea matters because numbers do not always act like amounts. A student number may contain digits, but it is only a label. A satisfaction score may use the numbers 1 to 5, but those numbers mainly show an order. Temperature in Celsius has equal steps, while weight has equal steps and a real zero.

The four levels of measurement in statistics are nominal, ordinal, interval, and ratio. Many books group nominal and ordinal variables as categorical variables. They group interval and ratio variables as quantitative or metric variables. Some courses use three broad levels by joining interval and ratio data under the metric level.

Why levels of measurement matter

Measurement scales guide nearly every step of a study. They affect how you write a survey question, store a variable, describe the results, choose a chart, and select a statistical test. A method that works well for one scale may give a meaningless result for another.

For example, you can count eye colours and report the most common colour. You cannot calculate a useful average eye colour. You can place satisfaction answers in order, but you should not automatically assume that the distance from poor to fair is exactly the same as the distance from good to excellent.

The level is not the only factor in analysis. The research question, sample size, distribution, study design, and assumptions also matter. Still, identifying the scale gives you a safe and logical starting point.

Before calculating anything, ask what the values truly represent. Are they labels, ranks, equal differences, or measured amounts?

Nominal level of measurement

Nominal data places observations into separate categories. The categories have names, but they do not have a natural order. One group is not above or below another group.

Examples include type of internet connection, favourite school subject, country of residence, payment method, blood group, and transport type. A gender variable is usually nominal when its values are categories because those categories do not form a statistical ranking.

A nominal variable with only two categories is called a binary or dichotomous variable. Examples include present or absent, yes or no, and passed or not passed.

Useful summaries for nominal data

• Absolute frequency and percentage 
• Mode, which is the most common category
• Frequency table
• Bar chartPie
• chart when a part of a whole comparison is useful

Suppose 20 students report their preferred study location. Eight choose the library, six choose home, four choose a cafe, and two choose a study hall. The library is the mode. The mean and median are not meaningful because the categories are names, not measured values.

 

Ordinal level of measurement

Ordinal data also uses categories, but the categories have a meaningful order. You can tell which answer is higher, lower, earlier, later, better, or worse. The exact distance between two positions is unknown or may not be equal.

Examples include customer satisfaction from very dissatisfied to very satisfied, competition positions, education level, pain severity, product quality, and agreement choices. First place ranks above second place, but the performance gap between first and second may be much smaller than the gap between second and third.

For ordinal data, frequencies, percentages, the mode, the median, and percentiles may be useful. An ordered bar chart keeps the natural ranking clear. Researchers sometimes calculate a mean for a rating scale, but that choice needs judgment and a clear explanation of the assumptions.

Categorical variables and number codes

Nominal and ordinal variables are both categorical variables. Nominal categories have no order. Ordinal categories do have an order. A number code does not change the measurement level.

For example, a company may store department codes as 1 for finance, 2 for sales, and 3 for support. The codes are still nominal. Department 3 is not three times department 1, and an average department code has no practical meaning.

Interval level of measurement

Interval data has a meaningful order and equal distances between values. A five unit change means the same amount anywhere on the scale. However, the scale does not have a true zero.

A true zero means that none of the measured quantity is present. Zero degrees Celsius does not mean that temperature is absent. It is simply a point on the Celsius scale. This is why 20 degrees Celsius is not twice as hot as 10 degrees Celsius, even though the difference between 10 and 20 is the same as the difference between 20 and 30.

Calendar years are another common example. The difference between 2010 and 2020 is ten years, just like the difference between 2020 and 2030. Yet the year zero does not represent the absence of time. Some standardised scores may also be treated as interval data when equal differences are supported by the scale design.

Ratio level of measurement

Ratio data has order, equal intervals, and a true zero. Zero represents the absence of the measured amount. Because of this, ratio statements are meaningful.

Height, weight, age, distance, time taken, income, number of purchases, and electricity use are common examples. Zero minutes of exercise means no exercise time. A weight of 10 kilograms is twice a weight of 5 kilograms.

Age is normally a ratio variable when it is measured as an exact amount of time since birth. If age is changed into groups such as child, adult, and older adult, it becomes ordinal because the groups have an order but do not preserve exact equal gaps.

Interval versus ratio data

Both interval and ratio scales have ordered values and equal gaps. The key difference is the zero point.

Nominal, ordinal, interval, and ratio comparison

The table below compares the four measurement scales and the information each one provides.

Levels of measurement in a student survey

A college wants to understand study habits. The survey includes several variables, and each variable carries a different type of information.

Common mistakes to avoid

1. Assuming every number is quantitative. Telephone numbers, postal codes, shirt numbers, and identification numbers contain digits but usually act as nominal labels.

2. azcCalculating a mean for nominal categories. Counts and percentages can describe transport types, but an average transport type has no meaning.

3. Assuming equal gaps in every rating scale. Ordered answers do not automatically prove that each step has the same size.

4. Confusing interval and ratio data. Both have equal intervals, but only ratio data has a true zero.

5. Treating category codes as measured amounts. Codes make data storage easier, but they do not create real quantities.

6. Choosing a chart that does not match the variable. Bars are usually clearer for categories, while histograms are designed for measured numerical values.

Levels of measurement and statistical analysis

Nominal data is often described with counts, percentages, proportions, and the mode. Relationships between nominal variables may be studied with methods based on frequency tables, depending on the design and assumptions.

Ordinal data can use the median, percentiles, ranks, and rank based methods. Because the exact gaps are not guaranteed to be equal, methods that preserve order without assuming equal distance can be useful.

Interval and ratio data often allow means, standard deviations, correlations, regression models, and many common statistical tests. The correct method still depends on the research question, sample size, shape of the distribution, independence of observations, and other assumptions.

Levels of measurement and charts

Nominal data often works well in a bar chart because the chart compares separate categories. Ordinal data also works well in an ordered bar chart, where the category order is kept from low to high.

Interval and ratio data may be displayed with a histogram, box plot, scatter plot, or line chart. A histogram shows the shape of one numerical distribution. A scatter plot shows the relationship between two numerical variables. A line chart is useful when values are measured across time or another natural sequence.

Metric variables

Interval and ratio variables may be called metric variables because differences between values are meaningful. Ratio variables also support meaningful ratios because they have a true zero.

A numerical appearance does not make a variable metric. A postal code, telephone number, or employee number should usually remain nominal because the digits identify a category or person rather than measure an amount.

Frequently asked questions

What are the four levels of measurement?

Why is the level of measurement important?

What is the difference between nominal and ordinal data?

What is the difference between interval and ratio data?

Is a Likert scale ordinal or interval?

Are all numerical variables interval or ratio variables?

Is age ratio or interval data?

Is gender nominal or ordinal data?

Summary

The level of measurement shapes the story your data can tell. Nominal data names. Ordinal data ranks. Interval data measures equal differences without a true zero. Ratio data measures equal differences with a true zero. Once you know the scale, you can choose clearer charts, more suitable summaries, and better statistical methods.

Suggested page slug: level-of-measurement
Main search topic: levels of measurement in statistics
Related search language used naturally on this page: nominal ordinal interval ratio, nominal versus ordinal, interval versus ratio, Likert scale ordinal or interval, age ratio or interval, and examples of levels of measurement