Levels of measurement

Measurement is the process of assigning numbers to variables or events according to rules. Every variable in a study that is assigned a specific number must be similar to every other variable assigned that number. In other words, if you decide to assign the number 1 to represent all male subjects in a sample and the number 2 to represent all female subjects, you must use this same numbering scheme throughout your study.

The measurement level is determined by the nature of the object or event being measured. Understanding the levels of measurement is an important first step when you evaluate the statistical analyses used in a study. There are four levels of measurement: nominal, ordinal, interval, and ratio (Table 16-1). The level of measurement of each variable determines the type of statistic that can be used to answer a research question or test a hypothesis. The higher the level of measurement, the greater the flexibility the researcher has in choosing statistical procedures. Every attempt should be made to use the highest level of measurement possible so that the maximum amount of information will be obtained from the data. The following Critical Thinking Decision Path illustrates the relationship between levels of measurement and appropriate choice of descriptive statistics.

CRITICAL THINKING DECISION PATH

Descriptive Statistics

Nominal measurement is used to classify variables or events into categories. The categories are mutually exclusive; the variable or event either has or does not have the characteristic. The numbers assigned to each category are only labels; such numbers do not indicate more or less of a characteristic. Nominal-level measurement can be used to categorize a sample on such information as gender, marital status, or religious affiliation. For example, Alhusen and colleagues (2012; Appendix B) measured marital status using a nominal level of measurement. Nominal-level measurement is the lowest level and allows for the least amount of statistical manipulation. When using nominal-level variables, typically the frequency and percent are calculated. For example, Alhusen and colleagues (2012) found that among their sample of pregnant women 54% were single, 34% were partnered, and 10% were married.

A variable at the nominal level can also be categorized as either a dichotomous or categorical variable. A dichotomous (nominal) variable is one that has only two true values, such as true/false or yes/no. For example, in the Thomas and colleagues (2012; Appendix A) study the variable gender (male/female) is dichotomous because it has only two possible values. On the other hand, nominal variables that are categorical still have mutually exclusive categories but have more than two true values, such as marital status in the Alhusen and colleagues study (single, partnered/not married, married, other).

Ordinal measurement is used to show relative rankings of variables or events. The numbers assigned to each category can be compared, and a member of a higher category can be said to have more of an attribute than a person in a lower category. The intervals between numbers on the scale are not necessarily equal, and there is no absolute zero. For example, ordinal measurement is used to formulate class rankings, where one student can be ranked higher or lower than another. However, the difference in actual grade point average between students may differ widely. Another example is ranking individuals by their level of wellness or by their ability to carry out activities of daily living. Alhusen and colleagues used an ordinal variable to measure the household income of pregnant women in their study and found that 46% (n = 76) of the study participants had household incomes under $10,000. Ordinal-level data are limited in the amount of mathematical manipulation possible. Frequencies, percentages, medians, percentiles, and rank order coefficients of correlation can be calculated for ordinal-level data.

Interval measurement shows rankings of events or variables on a scale with equal intervals between the numbers. The zero point remains arbitrary and not absolute. For example, interval measurements are used in measuring temperatures on the Fahrenheit scale. The distances between degrees are equal, but the zero point is arbitrary and does not represent the absence of temperature. Test scores also represent interval data. The differences between test scores represent equal intervals, but a zero does not represent the total absence of knowledge.

In many areas in the social sciences, including nursing, the classification of the level of measurement of scales that use Likert-type response options to measure concepts such as quality of life, depression, functional status, or social support is controversial, with some regarding these measurements as ordinal and others as interval. You need to be aware of this controversy and look at each study individually in terms of how the data are analyzed. Interval-level data allow more manipulation of data, including the addition and subtraction of numbers and the calculation of means. This additional manipulation is why many argue for classifying behavioral scale data as interval level. For example, Alhusen and colleagues (2012) used the Maternal-Fetal Attachment Scale (MFAS), which rates 24 items related to attachment on a 5-point Likert scale from 1 (definitely no) to 5 (definitely yes) with higher scores indicating greater maternal-fetal attachment. They reported the mean MFAS score as 84.1 with a standard deviation (SD) of 14.2.

Ratio measurement shows rankings of events or variables on scales with equal intervals and absolute zeros. The number represents the actual amount of the property the object possesses. Ratio measurement is the highest level of measurement, but it is most often used in the physical sciences. Examples of ratio-level data that are commonly used in nursing research are height, weight, pulse, and blood pressure. All mathematical procedures can be performed on data from ratio scales. Therefore the use of any statistical procedure is possible as long as it is appropriate to the design of the study.

One way of organizing descriptive data is by using a frequency distribution. In a frequency distribution the number of times each event occurs is counted. The data can also be grouped and the frequency of each group reported. Table 16-2 shows the results of an examination given to a class of 51 students. The results of the examination are reported in several ways. The columns on the left give the raw data tally and the frequency for each grade, and the columns on the right give the grouped data tally and grouped frequencies.

When data are grouped, it is necessary to define the size of the group or the interval width so that no score will fall into two groups and each group will be mutually exclusive. The grouping of the data in Table 16-2 prevents overlap; each score falls into only one group. The grouping should allow for a precise presentation of the data without serious loss of information.

Information about frequency distributions may be presented in the form of a table, such as Table 16-2, or in graphic form. Figure 16-1 illustrates the most common graphic forms: the histogram and the frequency polygon. The two graphic methods are similar in that both plot scores, or percentages of occurrence, against frequency. The greater the number of points plotted, the smoother the resulting graph. The shape of the resulting graph allows for observations that further describe the data.

Measures of central tendency

Measures of central tendency are used to describe the pattern of responses among a sample. Measures of central tendency include the mean, median, and mode. They yield a single number that describes the middle of the group and summarize the members of a sample. Each measure of central tendency has a specific use and is most appropriate to specific kinds of measurement and types of distributions.

The mean is the arithmetical average of all the scores (add all of the values in a distribution and divide by the total number of values) and is used with interval or ratio data. The mean is the most widely used measure of central tendency. Most statistical tests of significance use the mean. The mean is affected by every score and can change greatly with extreme scores, especially in studies that have a limited sample size. The mean is generally considered the single best point for summarizing data when using interval- or ratio-level data. You can find the mean in research reports by looking for the symbols M = or .

The median is the score where 50% of the scores are above it and 50% of the scores are below it. The median is not sensitive to extremes in high and low scores. It is best used when the data are skewed (see Normal Distribution in this chapter), and the researcher is interested in the “typical” score. For example, if age is a variable and there is a wide range with extreme scores that may affect the mean, it would be appropriate to also report the median. The median is easy to find either by inspection or by calculation and can be used with ordinal-, interval-, and ratio-level data.

The mode is the most frequent value in a distribution. The mode is determined by inspection of the frequency distribution (not by mathematical calculation). For example, in Table 16-2 the mode would be a score of 72 because nine students received this score and it represents the score that was attained by the greatest number of students. It is important to note that a sample distribution can have more than one mode. The number of modes contained in a distribution is called the modality of the distribution. It is also possible to have no mode when all scores in a distribution are different. The mode is most often used with nominal data but can be used with all levels of measurement. The mode cannot be used for calculations, and it is unstable; that is, the mode can fluctuate widely from sample to sample from the same population.

HELPFUL HINT

Of the three measures of central tendency, the mean is the most stable, the least affected by extremes, and the most useful for other calculations. The mean can only be calculated with interval and ratio data.

Only gold members can continue reading. Log In or Register to continue

Share this:

• Click to share on Twitter (Opens in new window)
• Click to share on Facebook (Opens in new window)

Related

Related posts:

1. Nonexperimental designs
2. Systematic reviews and clinical practice guidelines
3. Understanding research findings
4. Reliability and validity

Stay updated, free articles. Join our Telegram channel

Join