Scales

Classical Measurement Theory

Developed during the 1920s, classical measurement theory is currently the most frequently used theory for instrument development and validation. It is based on the true-score model developed by Charles Spearman in 1904 and consists of two theoretical concepts, namely, true scores and error scores. These concepts are theoretical because it is impossible to obtain the absolute true score or the absolute error score. However, it is possible to say that a true score is that which reflects what the person is actually experiencing and that an error score is the gap between actual experience and what is perceived as that experience. Any observed score (O) is therefore equal to the true score (T) plus the error score (E) and can be presented in the form of the following equation:

$$O=T+E.$$

According to classical measurement theory, reliability is based on the amount of error in an observed score for an individual. If the amount of error is quite small, reliability can be claimed. If however, the error is quite large, the scale is unreliable. Part of classical measurement theory is the domain-sampling model. According to this model, any particular scale can be composed of responses to a random sample of items from a hypothetical [Page 1455]domain of items. The purpose of any particular scale will be to estimate the scale that would be obtained if one could employ all the items in the domain. The score a subject would obtain if it were possible to test the whole domain is referred to as the true score. A sample of items is reliable to the extent that the score it produces correlates highly with the true score.

Scale Development

A scale must always be developed within a very specific theoretical framework, as the framework guides item development for the scale. Using the theoretical framework, an operational definition of the construct must be developed to guide the scale developer in the design of the specific items that will measure the construct. The domain sampling model of measurement—that there is an infinite pool of possible items that can measure a construct—is then used to develop the items. The skill lies in choosing the specific items that will lead to high content validity, that is, doing a good job of representing the domain that the researcher is trying to measure. The list is typically developed by writing down one attribute of the defined construct and then writing an item based on that attribute; these two steps are repeated until the required number of items has been generated.

...