Related Experiment Videos
The contribution of individual variables to Hotelling's T2, Wilks' lambda, and R2
1Department of Statistics, Brigham Young University, Provo, Utah 84602.
Biometrics
|June 1, 1993
Summary
This study analyzes how individual variables impact multivariate statistics like Hotelling's T2 and Wilks' lambda. Findings reveal variable effects depend on correlations and predictive contributions, influencing hypothesis testing outcomes.
Area of Science:
- Multivariate statistics
- Statistical modeling
- Hypothesis testing
Background:
- Understanding variable influence is crucial in multivariate analysis.
- Existing methods may not fully elucidate individual predictor impacts.
- Hotelling's T2, Wilks' lambda, and R2 are key metrics in statistical inference.
Purpose of the Study:
- To quantify the effect of individual variables on Hotelling's T2, Wilks' lambda, and R2.
- To identify factors driving these effects in multivariate and regression analyses.
- To provide insights into variable contributions for hypothesis falsification.
Main Methods:
- Analysis of one-sample and two-sample Hotelling's T2 statistics.
- Evaluation of Wilks' lambda for multivariate analysis of variance (MANOVA).
- Assessment of R2 in multiple regression models.
Main Results:
- For Hotelling's T2, each variable increases the statistic, influenced by its multiple correlation with other variables and its predictive contribution.
- The effect of predictor variables on R2 mirrors their effect on T2.
- For Wilks' lambda, variables decrease the statistic, dependent on their individual F-statistic and changes in multiple correlation.
Conclusions:
- Individual variable effects on multivariate statistics are quantifiable and predictable.
- Understanding these effects aids in interpreting hypothesis testing results.
- The study offers a framework for assessing variable importance in complex statistical models.