Multivariate relative importance: extending relative weight analysis to multivariate criterion spaces

Area of Science:

• Organizational behavior
• Quantitative psychology
• Statistical modeling

Background:

• Evaluating predictor importance in regression is crucial for organizational scholars.
• Existing methods like general dominance weights and relative weights show promise but have limitations with multidimensional criteria.
• Investigating predictor importance with multiple correlated criteria requires advanced statistical techniques.

Purpose of the Study:

• To extend the understanding and application of relative importance statistics to multivariate regression designs.
• To introduce a novel procedure for estimating predictor importance when dealing with multiple correlated criterion variables.
• To provide an intuitive index for assessing the relationship between predictors and criteria in complex models.

Main Methods:

• Review of the concept of relative importance in statistical analysis.
• Development and discussion of a new procedure for calculating multivariate relative weights.
• Reanalysis of a published correlation matrix using the proposed method.
• Conducting a Monte Carlo simulation to compare the new procedure with existing techniques.

Main Results:

• The proposed multivariate relative weights provide an interpretable index of predictor-criterion relationships.
• Canonical solutions for multivariate importance are often uninterpretable.
• The new procedure offers a more intuitive understanding compared to other techniques.

Conclusions:

• The multivariate relative weights procedure effectively addresses the challenge of assessing predictor importance with multiple criteria.
• This method offers significant implications for organizational researchers by providing clearer insights into complex relationships.
• The findings suggest a more intuitive and reliable approach to evaluating predictor importance in multivariate contexts.