Description
A three-part rule for pruning scale items after exploratory factor analysis: primary loading above .40, all other loadings below .30, and a gap of at least .20.
When a questionnaire is factor analysed, each item should represent one factor clearly. Items that load on several factors, or on none, blur the meaning of a scale: cross-loading items carry variance from another construct into the scale score and inflate correlations between scales. The literature offers many separate cut-offs (primary loadings from .30 to above .45, cross-loading ceilings of .32 or .40, primary-secondary gaps of .20). Howard combined them into a single explicit rule so that retention decisions are consistent and reportable. His audit of HCI and cyberpsychology articles found that nearly half stated no primary cut-off and about four in five never mentioned cross-loadings, so stating all three thresholds is itself part of the rigour.
Attributes
- Discipline
- Survey & Psychometrics
- Loss type
- Siloed
- Stage
- Discovery, Evaluation
- Data
- Quantitative
- Effort
- Days
Steps
- Run the exploratory factor analysis through extraction, factor retention and rotation, and output the full pattern matrix for every item.
- Check that each item's loading on its own (primary) factor exceeds .40.
- Check that every loading on the other factors is below .30.
- Check that the primary loading exceeds the largest alternative loading by at least .20.
- Flag items failing any of the three conditions, review their wording for double-barrelled or ambiguous content, and remove them.
- Re-run the analysis on the reduced item set to confirm the structure holds and no new cross-loadings appear.
- Report all three cut-offs, the items removed and the final loadings.
Sources
- A Review of Exploratory Factor Analysis (EFA) Decisions and Overview of Current Practices (Howard, 2016)
Also known as
- .40-.30-.20 rule
- Cross-loading cut-off
- Primary and secondary loading criteria
- Item retention by factor loadings
- Factor loading and cross-loading cut-off rule for item retention
Categories
- Analytics & Behaviour
- Survey & Psychometrics