Intervalskala

An intervalskala, or interval scale, is a level of measurement in statistics where the order of values is meaningful and the differences between values are equal across the scale. Unlike a ratio scale, an interval scale has an arbitrary zero point, so ratios between values are not interpretable. On an interval scale you can talk about the distance between values and you can add or subtract values, but you cannot meaningfully multiply or divide to obtain a ratio.

Properties of an interval scale include equal intervals between adjacent values, a meaningful ordering, and a non-absolute zero. These properties allow for descriptive statistics such as the mean and standard deviation, as well as analyses that rely on differences, such as t-tests and ANOVA. Transformations that preserve the interval nature are affine: x' = a + b x with b > 0. This means the scale can be shifted or rescaled without changing the interpretation of differences.

Common examples include the Celsius and Fahrenheit temperature scales, calendar years, and many standardized test scores. In analysis, interval data support arithmetic on differences and central tendency measures, and enable Pearson correlation and linear regression under the assumption that the data are at least interval.

Limitations include the inapplicability of meaningful ratio statements (for example, one value being twice another) due to the arbitrary zero. It is also important to ensure that data considered interval genuinely meet that level of measurement; in practice some instruments or aggregated scores are debated as truly interval, and treating ordinal data as interval should be done with caution.