A New Perspective On Sequential Testing Procedures In Canonical Analysis: A Monte Carlo Evaluation.

Area of Science:

• Multivariate statistics
• Statistical hypothesis testing

Background:

• Canonical analysis is a statistical technique used to assess relationships between two sets of variables.
• Determining the number of significant canonical roots is crucial for accurate interpretation.
• Sequential testing procedures have faced criticism regarding their validity in this context.

Purpose of the Study:

• To compare the effectiveness of four distinct testing procedures for identifying the number of nonzero population roots in canonical analysis.
• To evaluate the validity of sequential testing methods against a proposed alternative (Harris method).

Main Methods:

• A Monte Carlo simulation study was employed to generate data and test the procedures.
• Four methods were compared: three sequential testing approaches and one advocated by Harris (1976).

Main Results:

• The study found the sequential testing procedures to be effective in establishing the number of nonzero roots.
• The Harris method, in contrast, demonstrated relative ineffectiveness.
• Results indicate sequential methods perform reliably.

Conclusions:

• Criticisms leveled against sequential testing for nonzero roots in canonical analysis appear unfounded.
• Sequential testing procedures remain a valid and effective approach for this statistical task.