Constrained Dual Scaling for Detecting Response Styles in Categorical Data

Area of Science:

• Multivariate statistics
• Psychometrics
• Data analysis

Background:

• Dual scaling (DS) is a multivariate exploratory technique.
• DS is equivalent to correspondence analysis for contingency tables.
• Discrepancies exist in DS and correspondence analysis for rating data.

Purpose of the Study:

• Exploit a DS peculiarity to detect response styles in rating data.
• Introduce a method to identify and analyze response styles.
• Improve the accuracy of data analysis by accounting for response biases.

Main Methods:

• Developed a spline-based constrained version of Dual Scaling (DS).
• The method detects four prominent types of response styles.
• Extended the method to accommodate multiple response styles.
• Employed an alternating nonnegative least squares algorithm for parameter estimation.

Main Results:

• Demonstrated that a DS peculiarity can be used to detect response styles.
• The proposed spline-based constrained DS method successfully identifies response style presence.
• Simulation studies and an empirical application validated the method's effectiveness.

Conclusions:

• The novel DS approach effectively detects and analyzes response styles in rating data.
• Accounting for response styles enhances the reliability of multivariate data analysis.
• This method offers a valuable tool for researchers analyzing rating scale data.