Description
Build a scale by keeping the items whose answers best separate known criterion groups, and add validity items that flag lying, faking or careless answering.
Empirical criterion keying selects items by what they predict rather than what they appear to measure. The MMPI was built by writing more than 1,000 true-false statements, giving them to people with known disorders and to comparison groups, and keeping the items that best differentiated the groups. Kept items were combined into subscales (the MMPI-2 has more than 500 items, with clinical scales such as depression at 57 items and social introversion at 69), and interpretation rests on the respondent's profile across subscales. The inventory also includes validity items designed to detect lying, faking and non-response. Because items are chosen statistically, their content may not obviously match the construct, which makes them harder to fake but also harder to explain. It is a distinct strategy from factor-analytic scale development, which groups items by their intercorrelations.
Attributes
- Discipline
- Survey & Psychometrics
- Loss type
- Siloed
- Stage
- Discovery, Definition
- Data
- Quantitative
- Effort
- Weeks
Steps
- Write a large pool of candidate items.
- Administer them to people known to belong to each criterion group (for example churned versus retained customers) and to a comparison group.
- Keep the items whose answers best differentiate the groups, and cross-validate the selection on a fresh sample.
- Combine kept items into subscales.
- Add validity items to detect lying, faking, inattention and non-response.
- Interpret respondents by their profile across subscales, together with their validity flags.
Sources
- MIT Psychology (Stangor, 2012)
- SUS: A quick and dirty usability scale (John Brooke)
Also known as
- Contrasted-groups item selection
- MMPI-style item selection
- Criterion-keyed scale construction
- Contrasted-groups method
Categories
- Analytics & Behaviour
- Survey & Psychometrics