Measurement in Statistics: Data Types and Levels of Measurement

Data Types

• Data are broadly classified into two primary categories: qualitative and quantitative.

  • Qualitative (or categorical) data consist of values that can be placed into nonnumerical categories. These represent qualities or labels rather than numerical magnitudes.

  • Quantitative data consist of values representing counts or measurements. These are numerical in nature and allow for mathematical comparisons.

• Example 1: Classifying Data Types

  • Brand names of shoes in a consumer survey: Classified as qualitative because brand names are categorical and do not represent a numerical count or measurement.

  • Scores on a multiple-choice exam: Classified as quantitative because they represent a specific count of the number of correct answers.

Discrete versus Continuous Data

• In the context of quantitative data, values can be further subdivided into discrete or continuous categories.

  • Continuous data can take on any value within a given interval. There are no gaps between possible values; between any two numbers, another number can always exist.

  • Discrete data can take on only particular, distinct values and not other values in between. These often result from counting whole items.

• Example 2: Discrete or Continuous?

  • Measurements of the time it takes to walk a mile: These represent continuous data because time can take on any value (including fractions and decimals) within an interval.

  • The number of calendar years (such as $2007, 2008, 2009$): These represent discrete data because years change in distinct increments. For instance, on New Year’s Eve of $2009$, the year changes immediately from $2009$ to $2010$; we do not refer to fractional years in this context.

Levels of Measurement

There are four distinct levels of measurement used to classify data: nominal, ordinal, interval, and ratio.

• Nominal Level of Measurement

  • This is the simplest level of measurement.

  • It applies to variables that are described solely by names, labels, or categories.

  • The data consist of names, labels, or categories only and are strictly qualitative.

  • The data cannot be ranked or ordered in a meaningful way.

  • Example: Numbers on uniforms that identify players on a basketball team are at the nominal level. These numbers act merely as labels for identification and do not imply any rank or relative value.

• Ordinal Level of Measurement

  • This level applies to qualitative data that can be arranged in a meaningful order or ranking scheme (such as from low to high).

  • While the data can be ranked, we cannot determine precise or meaningful differences between those rankings.

  • It generally does not make sense to perform mathematical computations with data at this level.

  • Example: Student rankings of cafeteria food as "excellent," "good," "fair," or "poor" are at the ordinal level because there is a definite order to the categories, but the "distance" between "good" and "fair" is not numerically defined.

• Interval Level of Measurement

  • This level applies to quantitative data where intervals (the difference between values) are meaningful.

  • However, ratios are not meaningful because the data lack a true zero point. A value of zero at this level is arbitrary and does not represent the complete absence of the quantity.

  • Example 1: Calendar years of historic events (e.g., $1776, 1945, 2001$). An interval of one year always has the same meaning ($365$ or $366$ days). However, the year $0$ is arbitrary and does not mean "the beginning of time."

  • Example 2: Temperatures on the Celsius or Fahrenheit scales. An interval of $1^{\circ}C$ always has the same meaning. However, $0^{\circ}C$ (the freezing point of water) is an arbitrary zero and does not mean "no heat."

• Ratio Level of Measurement

  • This level applies to quantitative data where both intervals and ratios are meaningful.

  • Data at this level have a true zero point, meaning a value of $0$ represents a complete absence of the characteristic being measured.

  • Because of the true zero, we can say that one value is a certain multiple (e.g., twice as much or half as much) of another.

  • Example 1: Runners' times in the Boston Marathon. A time of $6\,\text{hours}$ is exactly twice as long as a time of $3\,\text{hours}$ because a time of $0\,\text{hours}$ represents the true starting point (the absence of time elapsed).

  • Example 2: The Kelvin Temperature Scale. Scientists use the Kelvin scale because it has a true zero called absolute zero. A temperature of $0\,K$ is the coldest possible temperature, equivalent to approximately $-273.15^{\circ}C$ or $-459.67^{\circ}F$. Note that the degree symbol is not used for Kelvin temperatures (e.g., $0\,K$ not $0^{\circ}K$).

Think About It: The Impact of Coding Nominal Data

• Scenario: Consider a survey asking for a favorite flavor of ice cream (e.g., vanilla, chocolate, cookies and cream). This data is at the nominal level.

• Problem: If a researcher assigns numbers to these flavors for computer entry (e.g., $1 = \text{vanilla}$, $2 = \text{chocolate}$, $3 = \text{cookies and cream}$), does this change the data to the ordinal level?

• Analysis: No, it does not. Assigning numbers for convenience does not create a meaningful ranking or order based on magnitude or quality. The numbers remain arbitrary labels, maintaining the nominal level of measurement.