Interval data sits one step below the ratio on the measurement hierarchy, it has ordered categories and equal gaps between them, but it lacks a true zero. That missing zero sounds like a minor technical detail, but it has real consequences for how you can interpret your results.
Defining Interval Measurement
There are two properties that mainly define interval data:
- Equal intervals – the difference between any two adjacent points on the scale represents the same quantity throughout the scale.
- No true zero – the zero point is arbitrary or conventional, not an absence of the measured attribute.
That second property is the key distinction from ratio data. Because the zero is arbitrary, you cannot form meaningful ratio statements. You can add and subtract values on an interval scale meaningfully, but you cannot multiply or divide them and expect sensible results.
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!
Classic Examples
| Variable | Level | Why Not Ratio? |
|---|---|---|
| Temperature in Celsius | Interval | 0°C is the freezing point of water, not ‘no temperature’ |
| Temperature in Fahrenheit | Interval | 0°F is equally arbitrary – based on a saltwater solution |
| Temperature in Kelvin | Ratio | 0K is absolute zero – complete absence of thermal energy |
| IQ score | Interval | IQ of 0 doesn’t mean ‘no intelligence’ – it’s outside the scale range |
| Calendar year (AD) | Interval | Year 0 is a convention, not the beginning of time |
| Standardised test score (mean=100) | Interval | Score of 0 means ‘no score on this test’, not ‘no ability’ |
| Reaction time (ms) | Ratio | 0ms = no time elapsed — genuine absence |
What Statistics Can You Use
Interval data unlocks the full range of parametric statistical methods. This is a significant jump from ordinal data, which is restricted to non-parametric alternatives.
Appropriate for Interval Data
- Arithmetic mean – equal intervals mean the mean is a meaningful summary statistic.
- Standard deviation and variance – measure how spread out data is around the mean.
- Pearson’s r – correlation between two interval (or ratio) variables.
- Independent samples t-test – compare means between two groups.
- Paired samples t-test – compare means from the same group at two time points.
- One-way ANOVA – compare means across three or more groups.
- Linear regression – predict an interval outcome from one or more predictors.
Not Appropriate for Interval Data
- Ratio statements – you cannot say 60°C is ‘twice as warm’ as 30°C.
- Coefficient of variation – this requires a meaningful zero to be interpretable.
Example: A study comparing maths attainment between state and independent school pupils in England uses KS2 standardised scores (treated as interval). The researcher calculates means and standard deviations for each group, then runs an independent sample t-test to determine whether the difference is statistically significant.
The Likert Scale Debate
Here is an issue that comes up in almost every research methods class. Likert scale items – strongly agree / agree / neutral / disagree / strongly disagree – are technically ordinal. The psychological gap between ‘agree’ and ‘neutral’ may not be the same as between ‘neutral’ and ‘disagree’.
Despite this, many researchers treat composite Likert scales (multiple items summed into a total score) as interval data, arguing that the summing process produces an approximately continuous, normally distributed outcome. This is a pragmatic position that enables more powerful statistics.
| Approach | Treats Likert as | Implications |
|---|---|---|
| Conservative / orthodox | Ordinal | Use median, non-parametric tests (Mann-Whitney U, Spearman’s rho) |
| Common in practice | Interval (composite scales) | Use mean, t-tests, ANOVA, Pearson’s r |
| Recommended | Check discipline norms | Be explicit about the assumption; cite methodological justification |
Student Tip: Ask your dissertation supervisor which convention your department follows before you commit to an analysis approach. Different disciplines have different norms, and using the wrong one for your field can raise unnecessary questions even if your analysis is otherwise sound.
Interval Vs Ratio In Practice
For most common statistical tests, t-tests, ANOVA, regression, Pearson’s correlation, the interval/ratio distinction does not change your choice of method. Both levels support the same parametric toolkit.
The distinction matters for interpretation. If your data is ratio, you can say ‘Group A’s mean was twice that of Group B.’ If it is interval, you can only say ‘Group A’s mean was 15 points higher.’ Precision in how you describe your findings reflects methodological literacy.
Frequently Asked Questions
Interval data is yet another type of data that can be calculated along a scale where every point is placed at an equal interval from another, just as the name explains itself.
The data collected on a thermometer is an example of interval data, as its gradation markings are equally distanced from each other. Interval data is always expressed in the form of numbers, unlike ordinal data.
Methods or techniques used by researchers depend on the data usage, the person collecting the data, and the audience being targeted.
Following are interval data collection techniques:
- Observations
- Surveys and Questionnaires
- Interviews
Ratio. A person who is 0 years old genuinely has been alive for zero years (approximately), and someone who is 40 is twice the age of someone who is 20. The zero is meaningful. Age is ratio, not interval.
Yes. Linear regression requires that the dependent variable is at least interval level. Interval-level outcomes are fine, as long as other assumptions (normality of residuals, homoscedasticity, etc.) are also met.
Report the mean and standard deviation: M = 72.4, SD = 8.6. For group comparisons, report the t-test or ANOVA result alongside means and SDs for each group. Avoid ratio language, stick to differences rather than multiples.
