Level of Measurement — Definition & Example
Level of measurement is a label for how much information a variable's values carry. It tells you whether you can only name categories, put them in order, subtract them, or divide them.
Under S. S. Stevens' four-level scheme, a scale is nominal if values support only equality (, ). It is ordinal if values also support order (, ). It is interval if differences are meaningful but the zero point is arbitrary, so only transformations with preserve it. It is ratio if it also has a true zero, so only with preserve it and ratios of values are meaningful. Each level has all the properties of the levels before it.
Example
Problem: A class survey records four variables for each student: favorite color (blue), class rank (3rd), outdoor temperature at lunch (80°F), and height (60 inches). Classify each variable's level of measurement.
Step 1: Favorite color only names a category. Blue is not more or less than red, so color is nominal.
Step 2: Class rank has a clear order: 1st comes before 3rd. The gap between 1st and 2nd need not equal the gap between 2nd and 3rd, so rank is ordinal.
Step 3: Temperature in °F has equal-sized degrees, so differences make sense, but 0°F does not mean "no heat", so ratios do not (the Interval Scale page shows why). Temperature in °F is interval.
Step 4: Height has a true zero, where 0 inches means no height. Ratios make sense, so a 60-inch student is twice as tall as a 30-inch toddler. Height is ratio.
Answer: Color is nominal, class rank is ordinal, temperature in °F is interval, and height is ratio.
Why It Matters
The level of a variable decides which summary you can use. The mode works at every level, and the median needs at least ordinal data. The mean is meaningful for interval and ratio data. AP Statistics sorts variables into categorical and quantitative; nominal and ordinal variables both count as categorical and are usually shown in bar graphs. Interval and ratio variables are grouped as quantitative and shown in histograms or dot plots.
Common Mistakes
Mistake: Treating numbers that are really labels as quantities, such as averaging jersey numbers or ZIP codes.
Correction: A number does not make a variable quantitative. Jersey number 24 is not twice jersey number 12. These values are nominal, so report counts or the mode, not a mean.
