Constructing Scales from Measured Variables
Scales are constructions of measurements or observations that are intended to convey information about characteristics related to some trait or characteristics. We use widely defined scales to measure distance such as the meter and kilometer system and the inch, foot, mile scale. In psychological measurement the focus is frequently not an empirical physical characteristic but rather a trait that cannot be directly observed. In these cases the trait may still be measured by using indicators such as self-reports and/or implied from behavior . Frequently psychological tools use a series of questions to collect indications about some trait or psychological characteristic. The responses to the series of questions is then combined to develop some measure of the underlying trait. Developing a scale usually involves the following steps.
- Write items based on a theory and the concepts - construct validity
- Field-test the scale with several hundred users
- Apply factor analysis principal component extraction to a pool of items - used to confirm item coherence
- Analyze the scale looking at item to scale-without the item correlations and some form of reliability statistic - examining item coherence.
- Examine issues of test-retest and form comparability if appropriate.
- Conduct evaluations for predictable power and other forms of validity
When assessing a scale the two core questions focus on whether the items in a scale are consistent as indicators or the trait. This is the issue of internal reliability. And whether the scale measures the trait that it claims to measure, which is the question validity and predictability.
Assessment of these core issues are address in the statistical sections of these web pages.
