Types Of Data

1. Core Concept Overview & Intuition

Statistics is the science of collecting, analyzing, interpreting, presenting, and organizing data. Every statistical investigation begins with a single foundational question: what kind of object is being measured? The answer determines which summaries, displays, and later procedures are legitimate. The classification of data into types is therefore the first act of rigor in any study, because a measurement carries not only a value but also a scale of admissible operations.

Data are first partitioned into two broad families. Qualitative (categorical) data record a quality, label, or attribute — a membership in a category rather than a magnitude. Quantitative (numerical) data record a quantity, so that arithmetic on the values is meaningful. Each family then splits once more. Qualitative data are nominal when the categories are merely distinct labels with no inherent order, and ordinal when the categories possess a meaningful, non-arbitrary ordering. Quantitative data are discrete when the values arise from counting and are separated by gaps, and continuous when the values arise from measurement and can, in principle, fall anywhere in an interval.

The intuition that unifies the scheme is the hierarchy of structure. A nominal label supports only the judgement "same or different." An ordinal label additionally supports "less than or greater than." A discrete count supports counting and exact arithmetic. A continuous measurement supports the full range of arithmetic and interval reasoning, subject only to the precision of the measuring instrument. Moving down this hierarchy adds structure; moving up it discards information. Misclassifying data — treating an ordinal code as a nominal label, or a continuous measurement as a count — either wastes information or licenses operations that the data never supported.

A useful diagnostic is to ask what a difference between two recorded values would mean. For nominal data the difference is undefined. For ordinal data a difference in rank is meaningful but its magnitude is not. For discrete data a difference is an exact count of units. For continuous data a difference is a measured interval whose meaningfulness is bounded by instrument precision. This single question resolves nearly every classification encountered in practice.

2. Mathematical Formalization & Core Definitions

> Definition (Data). Data are the recorded outcomes of observation or measurement, organized so that they may be analyzed, interpreted, and presented. Statistics is the science of collecting, analyzing, interpreting, presenting, and organizing such data.

> Definition (Qualitative / Categorical Data). Qualitative or categorical data assign each observational unit to one of a set of categories. The recorded value is a label, and the categories are distinguished by name rather than by magnitude.

> Definition (Nominal Data). Nominal data are qualitative data whose categories admit no natural ordering. Any two distinct categories are related only by equality or inequality; no category is intrinsically "greater" or "less" than another.

> Definition (Ordinal Data). Ordinal data are qualitative data whose categories possess a meaningful, non-arbitrary ordering. Categories may be ranked, but the numerical distance between consecutive ranks carries no quantitative meaning.

> Definition (Quantitative / Numerical Data). Quantitative or numerical data assign each observational unit a number representing a magnitude, so that arithmetic operations on the values are meaningful.

> Definition (Discrete Data). Discrete data are quantitative data whose possible values form a set of separated values, typically arising from counting. Between two adjacent admissible values there are no intermediate admissible values.

> Definition (Continuous Data). Continuous data are quantitative data whose possible values may, in principle, take any value within an interval, typically arising from measurement. Any recorded value is limited only by the precision of the measuring instrument.

The four types form a nested hierarchy of admissible structure. Let $x$ and $y$ denote two recorded values. The following table summarizes which comparisons and operations are supported at each level.

The hierarchy is strict: every operation admissible at a higher level is admissible at every level below it in the table, but not conversely. Nominal data support the fewest operations; continuous data support the most.