Data is often presented in bands. For example, age may be shown as Under 18, 18 to 24, 25 to 34, etc.
Banding can occur either at the time of data capture (e.g., people indicate they are in the band 18 to 24), or, as a form of data reduction performed when analyzing the data.
Although bands are ordered, the data is technically not ordinal. Its properties are a function:
- The underlying data has been banded. This data may be ordinal, interval, or ratio. For example, when age is banded, the resulting data is a limited form of ratio data (e.g., somebody who is 65 or more is at least three times as old as somebody aged Under 18).
- The cuts are used to define the bands. For example, where interest rates are banded in one percent bands (0.00 to 0.99%, 1.00% to 1.99%, 2.00% to 2.99%, etc.), the data is much closer to having ratio-scale properties than when banded in 10% bands.
Although statistical methods can be developed specifically for this type of data, in practice it is usually treated as ordinal or interval. The resulting data always has ordinal properties. If a large number of brands are created and none are particularly wide, it is common to treat banded data as having the same properties as the underlying data (i.e., interval or ratio).
Bands are also referred to as brackets, buckets, and ranges.
Please sign in to leave a comment.