Типы шкал: номинальная (категориальные данные), порядковая, интервальная (разрешены операции сложения) и абсолютная (разрешена операция умножения). Дискретные данные (counts), процентные данные.
Gries, page 175:
The variable class with an intermediate information value is that of ordinal values. Just like categorical variables, ordinal variables provide the information that entities coded with different values are different, but in addition they also allow for rank-ordering the entities on the basis of these values—however, they do not allow to precisely quantify the difference between the entities. One example would be the variable SYNTACTICCOMPLEXITY of NPs. It is probably fair to assume that English NPs can be ordered in terms of their complexity as follows:
(13)
pronominal NPs (e.g. he) < simple lexical NPs (e.g. the book) < non-clausally modified lexical NPs (e.g. the green book or the book on the table) < clausally modified NPs (the book I bought last week).
An ordinal coding of SYNTACTICCOMPLEXITY could now be this:
In this coding, higher numbers reflect higher degrees of complexity so NPs can be ranked according to their complexity. However, note that the numbers that represent different degrees of complexity do not necessarily reflect the exact differences between the different NPs. Just because a non-clausally modified NP is coded as “3” while a pronominal NP is coded as “1”, this does not mean that the former is exactly three times as complex as the latter. Another linguistic example would be the IDIOMATICITY of a VP, which may be distinguished as “high (completely idiomatic)” vs. “intermediate (metaphorical)” vs. “low (fully compositional)”. A final, non-linguistic example would be grades: a student with an A (4.0) is not necessarily 1 1/3 times as good as a student with a B (3.0) or four times as good as a student with a D (1.0).
The next variable class to be distinguished here is that of interval variables. Interval variables provide the information of ordinal variables, but also provide meaningful information about the difference between different values. We will not treat interval variables here in much detail but rather focus on the, for our purposes, more useful class of ratio variables (interval and ratio variables together are sometimes also referred to as continuous variables). Ratio variables provide the information of interval variables, but in addition they also have a clear definition of 0, an absolute, non-arbitrary zero point. For example, the syllabic length of an NP is a ratio variable because it makes sense to say that an NP with four syllables is twice as long as one with two syllables and four times as long as one with just one syllable. Other examples for ratio variables would be pitch frequencies in Hz, word frequencies in a corpus, number of clauses between two successive occurrences of two ditransitives in a corpus file, the reaction time towards a stimulus in milliseconds…
Gries, page 175:
The variable class with an intermediate information value is that of ordinal values. Just like categorical variables, ordinal variables provide the information that entities coded with different values are different, but in addition they also allow for rank-ordering the entities on the basis of these values—however, they do not allow to precisely quantify the difference between the entities. One example would be the variable SYNTACTICCOMPLEXITY of NPs. It is probably fair to assume that English NPs can be ordered in terms of their complexity as follows:
The next variable class to be distinguished here is that of interval variables. Interval variables provide the information of ordinal variables, but also provide meaningful information about the difference between different values. We will not treat interval variables here in much detail but rather focus on the, for our purposes, more useful class of ratio variables (interval and ratio variables together are sometimes also referred to as continuous variables). Ratio variables provide the information of interval variables, but in addition they also have a clear definition of 0, an absolute, non-arbitrary zero point. For example, the syllabic length of an NP is a ratio variable because it makes sense to say that an NP with four syllables is twice as long as one with two syllables and four times as long as one with just one syllable. Other examples for ratio variables would be pitch frequencies in Hz, word frequencies in a corpus, number of clauses between two successive occurrences of two ditransitives in a corpus file, the reaction time towards a stimulus in milliseconds…