Ratio data is the highest level of measurement, and the most informative. It supports the full range of statistical operations, allows meaningful ratio comparisons, and is what most people picture when they think about quantitative research. Physical measurements, economic data, and most biological variables fall here.
Understanding what makes ratio data special, and where it differs from interval data, will sharpen both your analysis and your ability to interpret findings critically.
The Three Properties Of Ratio Data
- Ordered – you can rank observations from lowest to highest.
- Equal intervals – the difference between 10 and 20 represents the same quantity as between 50 and 60.
- True zero – zero means the complete absence of the attribute being measured.
That true zero is what separates ratio from interval data, and it is what makes ratio statements (double, half, 30% more) mathematically meaningful.
Looking for statistical analysis help?
Research Prospect to the rescue then!
We have expert writers on our team who are skilled at helping students with their research across a variety of disciplines. Guaranteeing 100% satisfaction!
Examples Of Ratio Data
| Variable | True Zero | Ratio Statement Valid | Example |
|---|---|---|---|
| Height (cm) | Yes – 0cm = no height | Yes | Average UK male height: 175.3cm (NHS data) |
| Weight (kg) | Yes – 0kg = no mass | Yes | BMI calculations used in NHS health checks |
| Income (£) | Yes – £0 = no income | Yes | ONS ASHE median full-time earnings: £34,963 (2023) |
| Reaction time (ms) | Yes – 0ms = no elapsed time | Yes | Used in cognitive psychology research at UK universities |
| Number of GP visits | Yes – 0 = no visits | Yes | NHS Digital patient contact data |
| Temperature (Kelvin) | Yes – 0K = absolute zero | Yes | Used in physics; not in everyday UK weather reporting |
| Temperature (Celsius) | No – 0°C = water freezes | No | UK Met Office temperature data is interval, not ratio |
What Statistics Can You Use
Ratio data unlocks the full parametric toolkit, plus ratio-specific statistics that are not meaningful at lower levels.
All Parametric Statistics
- Mean, standard deviation, variance.
- t-tests (independent and paired), ANOVA, MANOVA.
- Pearson’s correlation and linear regression.
- All their extensions and variations.
Ratio-Specific Statistics
- Coefficient of Variation (CV = standard deviation / mean × 100%) – expresses spread relative to the mean as a percentage. Only meaningful when the mean has a genuine scale (i.e., when data is ratio).
- Geometric mean – used with highly skewed ratio data, particularly in biological and economic research.
- Ratio comparisons – stating that Group A earns 1.8 times as much as Group B, or that response times in Condition X were 40% slower.
Example: A health psychology study at a UK university measures daily step counts (via accelerometer) in two groups: a mindfulness intervention group and a waitlist control group. Step count is ratio data (0 steps = no movement). The researcher reports means and standard deviations for each group, runs an independent t-test, and reports the result including Cohen’s d as an effect size. Saying the intervention group walked ‘twice as many steps’ is statistically valid because step count is ratio-level.
Ratio Data & Skewness
Ratio data is often skewed, particularly when it involves counts, incomes, or time measures. A positively skewed distribution (long right tail) is common, most values are relatively low, with a few very high outliers pulling the mean upward. When this happens:
- Report median alongside mean (or instead of it) to give a clearer picture of the typical value.
- Consider log transformations to normalise the distribution before parametric analysis.
- Use IQR alongside standard deviation to describe spread.
- Check for outliers systematically and decide in advance whether to include, exclude, or transform them.
| Scenario | Recommended Approach |
|---|---|
| Ratio data, approximately normal | Report mean and SD. Use parametric tests. |
| Ratio data, positively skewed (e.g., income, reaction times) | Report median and IQR alongside mean. Consider log transform before parametric tests, or use non-parametric equivalents. |
| Count data (integers, often 0) | Consider Poisson regression or negative binomial regression instead of ordinary linear regression. |
| Time-to-event data (e.g., days until readmission) | Consider survival analysis methods (e.g., Kaplan-Meier, Cox regression). |
Student Tip: If you collect ratio data that is highly skewed, do not just run a t-test and hope for the best. Check your distribution first using histograms and skewness statistics in SPSS or R. If skewness is a problem, a log transformation often helps, and it is straightforward to implement and justify.
Frequently Asked Questions
- Interval scales do not have a true zero, whereas ratio scales do.
- Negative numbers have meaning in interval scales, whereas negative numbers do not hold any meaningful value in ratio scales.
- Interval data can only be expressed using addition or subtraction, whereas ratio data can be expressed using addition, subtraction, multiplication, and division.
Some common examples of ratio scales are age, distance, speed, and mass.
What are the four levels of measurement of data?
The four levels of measurement of data are nominal, ordinals, interval, and ratio data.
Ratio. Age 0 means zero time elapsed since birth, and someone who is 60 is twice the age of someone who is 30. Age has a meaningful zero and supports ratio statements.
Usually yes, if 0% means getting nothing right and 100% means getting everything right, both the zero and the intervals are meaningful. However, scaled or curved exam scores may not have a genuine zero, in which case interval is more appropriate.
CV = (SD / Mean) × 100%. It tells you how large the spread is relative to the mean. Useful when comparing variability across datasets with different units or different means, for example, comparing income variability in two different regions where mean incomes differ substantially. It only makes sense for ratio data.
