Why Levels of Measurement Matter
In the previous lesson, we learned that data can be qualitative (categories) or quantitative (numbers). But there's a deeper layer to this. Not all categories are equal, and not all numbers work the same way. The level of measurement tells you what you can and can't do with your data.
Getting this right matters because using the wrong math on the wrong type of data gives you nonsense results. For instance, if you assign numbers to jersey colors (red = 1, blue = 2, green = 3), calculating the "average color" as 2 doesn't mean the average color is blue. That calculation is meaningless.
There are four levels of measurement, and each one builds on the previous. Let's walk through them from simplest to most informative.
Level 1: Nominal
Nominal data is the simplest level. It consists of names, labels, or categories with no natural order. "Nominal" comes from the Latin word for "name," and that's exactly what this level does: it names things.
- Blood type: A, B, AB, O. There's no sense in which A is "more" than B.
- Favorite cuisine: Italian, Mexican, Japanese, Indian. No ranking implied.
- Eye color: Brown, blue, green, hazel. Just labels.
- Zip codes: Even though they look like numbers, 90210 is not "greater" than 10001 in any meaningful way. They're labels for locations.
What you can do with nominal data: Count how many items fall into each category (frequency). Find the most common category (mode). That's about it. Calculating an average of nominal data makes no sense.
Level 2: Ordinal
Ordinal data has categories that follow a natural order or ranking. You can say one value is higher or lower than another. However, the distances between values are not necessarily equal.
- T-shirt sizes: Small, Medium, Large, Extra-Large. There's a clear order from smallest to largest, but the difference in fabric between Small and Medium isn't necessarily the same as between Large and Extra-Large.
- Customer satisfaction: Very Unsatisfied, Unsatisfied, Neutral, Satisfied, Very Satisfied. Satisfied is better than Neutral, but is the gap between them the same as between Unsatisfied and Neutral? We can't be sure.
- Education level: High school, Associate's degree, Bachelor's degree, Master's degree, Doctorate. There's a clear progression, but the "distance" between each step varies.
- Race finishing positions: 1st place, 2nd place, 3rd place. First is better than second, but the time gap could be a fraction of a second or several minutes.
What you can do with ordinal data: Everything you can do with nominal (count, find the mode), plus you can rank items and find the middle value (median). But calculating a true average is tricky because the gaps between categories aren't guaranteed to be equal.
You rate a restaurant 4 out of 5 stars. Your friend rates it 2 out of 5. Is your experience exactly "twice as good"? Probably not. Star ratings have an order (5 is better than 4), but the psychological distance between 1 star and 2 stars might feel very different from the distance between 4 stars and 5 stars. This is the hallmark of ordinal data: order exists, but equal spacing does not.
Level 3: Interval
Interval data has order and equal spacing between values. The difference between 10 and 20 is the same as the difference between 40 and 50. However, interval data has no true zero point, which means ratios don't work.
- Temperature in Fahrenheit or Celsius: The difference between 30 and 40 degrees is the same as between 70 and 80 degrees. But 0 degrees doesn't mean "no temperature." And 80 degrees is not "twice as hot" as 40 degrees.
- Calendar years: The gap between 1990 and 2000 is the same as between 2010 and 2020 (10 years). But the year 0 is an arbitrary reference point, not a true absence of time.
- IQ scores: The difference between 100 and 110 is meant to be the same as between 120 and 130. But an IQ of 0 doesn't mean "no intelligence," and a score of 140 is not "twice as smart" as 70.
What you can do with interval data: Everything from the previous levels, plus you can calculate meaningful averages and measure exact differences between values. But you cannot make ratio statements like "twice as much" because there's no true zero.
The "No True Zero" Idea
This is the part that confuses most people, so let's spend a moment on it. A "true zero" means the complete absence of the thing being measured. Zero degrees Celsius doesn't mean there's no heat. It's just the temperature at which water freezes, which is an arbitrary choice. Zero degrees Fahrenheit is a different arbitrary point. Because the zero is artificial, saying "40 degrees is twice as warm as 20 degrees" doesn't hold up.
Level 4: Ratio
Ratio data has everything interval data has, order, equal spacing, plus a true zero point. When zero means "none of it," you've got ratio data. This is the most informative level of measurement.
- Weight: 0 pounds means no weight. 100 pounds is genuinely twice as heavy as 50 pounds.
- Height: 0 inches means no height. A person who is 6 feet tall is twice as tall as a 3-foot-tall child.
- Money in your bank account: $0 means you have no money. $200 is exactly twice as much as $100.
- Cooking measurements: 0 cups of flour means no flour at all. A recipe calling for 2 cups uses exactly twice as much as one calling for 1 cup.
- Temperature in Kelvin: 0 Kelvin is absolute zero, the complete absence of thermal energy. This makes Kelvin a ratio scale, even though Fahrenheit and Celsius are interval scales.
What you can do with ratio data: Everything. Count, rank, average, compare differences, and make meaningful ratio statements ("A is three times as heavy as B"). This is the most flexible level.
A recipe calls for 2 cups of flour and 1 cup of sugar. You can say you need twice as much flour as sugar because cup measurements have a true zero (0 cups = no ingredient). That's a ratio comparison. Now imagine you bake the cake at 350 degrees Fahrenheit. You can't say 350 degrees is "twice as hot" as 175 degrees, because Fahrenheit doesn't have a true zero. Same kitchen, same cake, but different levels of measurement at play.
How to Remember the Four Levels
Think of the levels as building blocks, with each level adding a new capability:
- Nominal: Names only. You can group and count.
- Ordinal: Names + order. You can rank.
- Interval: Names + order + equal spacing. You can measure exact differences.
- Ratio: Names + order + equal spacing + true zero. You can compare ratios.
A helpful memory trick: think of the phrase "N-O-I-R" (the French word for black). Nominal, Ordinal, Interval, Ratio. Each letter represents a level, going from least to most informative.
Why This Matters in Practice
Choosing the wrong analysis for your level of measurement leads to misleading results. Here are some common mistakes:
- Averaging zip codes. The average of zip codes 10001 and 90210 is 50105.5. This number is meaningless because zip codes are nominal.
- Averaging star ratings without caution. A restaurant with an "average rating of 3.7 stars" is common practice, but technically questionable since star ratings are ordinal and the gaps between stars may not be equal.
- Saying "twice as hot." Reporting that 80 degrees Fahrenheit is "twice as warm as 40 degrees" is incorrect because Fahrenheit is an interval scale.
The four levels of measurement, nominal, ordinal, interval, and ratio, tell you what kind of comparisons and calculations are valid for your data. Nominal data is just labels. Ordinal data adds order. Interval data adds equal spacing. Ratio data adds a true zero, allowing the fullest range of analysis. Before you analyze any data, identify its level of measurement first. It will save you from drawing false conclusions.