Читать книгу Elementary Regression Modeling - Roger A. Wojtkiewicz - Страница 23

Measures themselves have characteristics, and the characteristics of measures relevant to statistical analysis are called the level of measurement of a variable. These characteristics can be framed as questions: Does the variable categorize units, does the variable order units, and does the variable capture a meaningful numerical amount of a characteristic that a unit possesses?

A variable that only categorizes is at the nominal level, a variable that categorizes and orders is at the ordinal level, and a variable that categorizes, orders, and captures a meaningful numerical amount is at the interval level.

The nominal level of measurement provides the least amount of information since it does not allow us to rank groups or compare groups based on how much of a characteristic they might have. Ordinal adds more information in allowing for the ordering of categories. Interval adds more information by allowing for the amount of the characteristic to be considered. Be aware that the ratio level is a fourth level of measurement. The ratio level adds a true zero to the interval level, but having a true zero is not a major issue in the analysis of survey data using regression procedures since the focus is on differences in levels and not on the levels themselves. In this book, I use the term interval to refer to both interval- and ratio-level variables.