Likert or Rasch? Nothing is more applicable than good theory

Area of Science:

• Nursing Research
• Psychometrics
• Health Sciences

Background:

• Questionnaires are widely used in nursing research to measure concepts.
• Unidimensional scaling assumes questionnaire items indicate a single concept.
• Summative Likert scaling is a common procedure, relying on classical test theory assumptions.

Purpose of the Study:

• To discuss and compare two scaling procedures: summative Likert scaling and the Rasch model.
• To explain the underlying rationales of both classical test theory and item response theory in measurement.
• To highlight the differences in item weighting and respondent/item scaling between the two methods.

Main Methods:

• Discussion of summative Likert scaling procedure based on classical test theory.
• Explanation of the Rasch model as an alternative scaling procedure based on item response theory.
• Comparison of item weighting and the concept of 'difficulty' in the Rasch model versus equal weighting in Likert scaling.

Main Results:

• Likert scaling assumes all items are parallel and equally weighted.
• The Rasch model scales respondents and items on the same continuum, differentiating items by 'difficulty'.
• Item response theory, underpinning the Rasch model, posits response probability depends on item difficulty and person ability.

Conclusions:

• The Rasch model provides an alternative to Likert scaling in nursing research.
• The Rasch model offers a more sophisticated approach to unidimensional scaling by accounting for item difficulty.
• Understanding these scaling procedures and their theoretical bases is crucial for accurate measurement in nursing research.