Multidimensional scaling vs. factor analysis of Wechsler's intelligence scales.

Area of Science:

• Psychometrics
• Cognitive Psychology
• Statistical Analysis

Background:

• Understanding the structure of human abilities is crucial in psychometrics.
• Principal component analysis (PCA) and factor analysis (FA) are common methods for exploring these structures.
• Multidimensional scaling (MDS) offers alternative approaches to visualizing data relationships.

Purpose of the Study:

• To explore the relationships between principal component/factor analysis and metric/nonmetric multidimensional scaling.
• To compare the structural findings of these methods when applied to intelligence test data.

Main Methods:

• Applied principal component analysis, principal factor analysis, metric multidimensional scaling, and nonmetric multidimensional scaling.
• Utilized standardization data from the Wechsler Adult Intelligence Scale-Revised (WAIS-R), Wechsler Intelligence Scale for Children-Revised (WISC-R), and Wechsler Preschool and Primary Scale of Intelligence (WPPSI).

Main Results:

• A general component/factor was identified in PCA/FA for all instruments, lacking a direct counterpart in MDS dimensions.
• The second component/factor in PCA/FA and the first dimension in MDS distinguished Verbal from Performance subtests.
• The third component/factor in PCA/FA and the second dimension in MDS appeared to represent 'Freedom from Distractibility'.

Conclusions:

• MDS methods may remove the significant general factor identified by PCA/FA, potentially limiting their utility in human abilities research.
• The findings highlight differences in how PCA/FA and MDS represent the underlying structure of intelligence test data.
• Careful consideration of the method's properties is needed when investigating the structure of human abilities.