Research Concepts
Measures
Once you know what variables are selected for data collection, you still have to decide how to measure the varialbes.
For some variables the measurement is obvious (height or weight or gender for example). Other times your measurement strategy is more complex.
Variables can either be classified as quantitative variables or as attribute variables.
An attribute variable, sometimes also called a nominative scale, describes a characteristic of the subject. These variables are usually measured with a word or label instead of a number. For example, political affiliation is usually measured by the name of the political party or the word "independent." (The researcher would need to decide betweeen self-identification and a declared affiliation on a voter registration.)
A quantitative variable is one in which there is a natural numerical measurement. These variables can use scales that are ordinal (which measures "larger" or "smaller" without specifying specific sizes), interval (a scale that indicates an amount but has no "zero") or ratio (a scale that has a true zero and equal intervals, like the real numbers). Ratio scales are sometimes called "continuous" scales.
Class rank is an example of an ordinal scale. Most psychometric tests, such as IQ, are interval scales. Measurements like height, weight or annual income are examples of ratio scales.
Note that an attribute variable can be coded with a number (0="male" and 1="female"). This coding does not change the variable to quantitative,it simply replaces a word label with a numerical label.
Also a variable might be measured in more than one way. For example, you could ask a subject for their annual income as reported on their federal tax return (resulting in a ratio scale). Alternatively, you could ask which category best describes their annual income:
- $15,000 or less per year
- $1500.01-$30,000 per year
- $30,000.01-$45,000 per year
- $45,000.01-$60,000 per year
- more than $60,000 per year
This measurement strategy -- which is less intrusive and may result in more accurate data reporting by subjects -- uses an ordinal scale rather than a ratio scale.
It is of course essential that the measures be objective -- this is the essence of scientific research. Failure to have objective measures will inevitably result in bias and flawed results.
