Quantitative Data
Quantitative data is information recorded as numbers that represent amounts — a height of 172 centimeters, 14 website visits, a temperature of 21 degrees, a test score of 83 — so that magnitude, difference, and sometimes ratio carry meaning. Because the numbers behave arithmetically, such data can be summarized with means and spreads, plotted on axes, modeled with equations, and tested statistically, which is why it is the default material of experiments, surveys with scales, sensors, financial records, and nearly all machine learning.
What qualifies data as quantitative is that the numbers mean something numerically: the distance between values is interpretable, and operations such as subtraction, averaging, or comparison of size produce meaningful results. A column of house prices qualifies; a column of area codes does not, despite being numerals, because "the average area code" says nothing. Quantitative data comes in two broad kinds that other entries treat in depth: counts, which arise from tallying whole units, and measurements, which arise from reading off a continuous quantity with an instrument. Whether it also supports ratio statements ("twice as heavy") depends on its level of measurement, the distinction Stanley Smith Stevens formalized in 1946.
The great strength of quantitative data is that it lets claims be checked with explicit error: precision, uncertainty, and comparison can all be expressed in numbers. Francis Galton's 1880s measurements of heights of parents and children produced regression and correlation; Karl Pearson built the mathematical machinery around such measurements; and the later development of confidence intervals and hypothesis tests (Gosset, Fisher, Neyman, and Pearson in the early twentieth century) provided formal rules for reasoning from samples. A quantitative result can be reproduced from the same data with the same code, and a different team can collect new data and see whether the effect persists, which is the working definition of a replication.
Quantitative data powers prediction as much as description: regression, forecasting, and almost every supervised learning model take numeric features and produce numeric or probabilistic outputs. Manufacturers track defect counts and dimensions in control charts; clinical trials compare blood-pressure readings between arms; analysts forecast demand from sales histories. Even when the original source is a photograph or a sentence, the model sees it as arrays of numbers, so the practical work is often turning other kinds of information into quantities and then checking that the conversion preserved what mattered, including the scaling and units applied to each feature.
Numerical precision is easily mistaken for accuracy, and that is the field's most common pitfall: a number with many decimals can still be systematically wrong. A scale that reads 70.3724 kilograms but is miscalibrated by two kilograms is precise and inaccurate. Other pitfalls include averaging quantities that should not be averaged, treating numerals that merely label groups as amounts, reporting figures with more significant digits than the instrument supports, and assuming that the quantitative is automatically more objective than the qualitative, when the choice of what to measure and how encodes human judgment. Data that is rich in numbers can also be poor in context, which is why numbers and narrative are best used together.
Read quantitative data as information that has been expressed as amount: its numbers are only as meaningful as the measurement behind them, but when the measuring is sound it lets uncertainty, comparison, and prediction be stated precisely instead of merely argued.