Measurements and Scale Data

Scales in Research

Measurement And Scaling

The role of scales in research

If someone is asked: "on a scale of 1-10, did you like that new movie," their response is meaningless unless it is clear that 1 means 'I didn't like at all,' that 5 means 'I thought it was okay,' and 10 means 'I loved it.' On a scale of 1-100, 10 is a not very positive response, while on a scale of 1-10, 10 is a very positive response. Before he or she begins accumulating data, a savvy researcher must create an effective, scaled response designed to measure the information that will be amassed and that will yield meaningful results.

But simply assigning a value of 1-10 alone is not scaling. Statistical analysis is required to ensure that analyzing the population sample's responses are accurate (Trochim, 2006, General issues in scaling). Three major types of uni-dimensional scaling methods exist. Thurstone or Equal-Appearing Interval Scaling offers a list of statements, to which individuals can agree or disagree. Then the likelihood of certain responses is calculated, based upon typical profiles of respondents. The Myers-Brigg or Jungian personality groupings are often tested through this method. This can be useful, as a variety of 'types' can be measured, based upon the scaling. For example, persons with a specific type of personality profile might favor a specific pattern of yes and no statements (such as a depressive type of personality vs. A manic personality). The tendency to manifest more than type of personality can be measured at any one time. There is no single 'correct' or 'incorrect' answer that can skew the test in a particular direction -- tendencies are observed over a long list of questions.

Likert or Summative Scaling creates a scale rating based upon the degree to which subjects or agree or disagree with certain statements (Trochim, 2006, Likert Scaling). This technique is often used in consumer research, where respondents will be asked to rate how much they like or dislike aspects of a particular product or phenomenon on a numerical scale. It has the advantage of being able to analyze 'shades' of meaning, rather than solely encapsulate responses in terms of black and white.

Finally, Guttman or 'Cumulative' Scaling creates a one-dimensional continuum for a concept under study (Trochim 2006, Guttman scaling). Guttman scores usually list items according to increasing intensity, in terms of agreement or disagreement (Trochim 2006, Guttman scaling). "Essentially, we would like a set of items or statements so that a respondent who agrees with any specific question in the list will also agree with all previous questions. Put more formally, we would like to be able to predict item responses perfectly knowing only the total score for the respondent" (Trochim, 2006, Guttman scaling). For example, to evaluate the liberal attitudes of a respondent regarding immigration, there might be a list of statements spanning from the relatively mild, such as agreeing that it was acceptable to allow more immigrants to migrate to the U.S., to more strongly favorable, such as supporting one's child marrying an immigrant.