|
 |
||||||||
Scale levels
A distinction is made between the following scale levels:
Nominal scale: In a nominal scale, different numbers do not mean anything more than different attributes. Which attribute to assign to which number is not important, as the numbers do not stand for “more” or “less”, “bigger” or “smaller”.
Examples: Query on
the field of study (1 = mathematics, 2 = informatics, etc.) or gender (1 = male, 2 = female) in a questionnaire.
Analysis possibilities: Frequency distributions.
Ordinal scale:
In an ordinal scale, the numbers represent a rank order; however the scale does not provide any information as to the relations of the attributes in the rank order. The intervals between the values do not correlate with the intervals “in reality”.
Example: rank order of the best five students in a quiz.
Analysis possibilities: Frequency distributions, analysis of rank information.
Interval
scale: In an interval scale, the numbers represent equal increments of the attribute being measured, but there is no "real" zero
point.
A response scale
of a questionnaire, for example, ranging from 1 (does not correspond at all) to
7
(corresponds fully) is usually interpreted as an interval scale.
Analysis possibilities: Frequency distributions, analysis of rank information, differences, totals, mean values.
Ratio
scale: In a ratio scale, the ratio of the intervals between the attributes is relevant. By contrast to an interval scale, the scale has a zero point which has a meaningful interpretation.
Example:
Age, number of certain actions in a log file.
Analysis possibilities: Frequency distributions, analysis of rank information, differences, totals, mean values.
