Smoothed Quantiles for Type Test Statistics with Applications

Area of Science:

• Statistical modeling
• Hypothesis testing
• Data analysis

Background:

• Chi-square type test statistics are crucial for model goodness-of-fit.
• Real-world data often violates assumptions, leading to inaccurate distributions.
• Bootstrap and Monte Carlo methods approximate distributions but struggle with small samples or limited data.

Purpose of the Study:

• To develop a statistical learning method for more accurate quantile estimation of chi-square type test statistics.
• To provide formulas for smoothing these quantiles.
• To apply smoothed quantiles in equivalence testing for mean and covariance structures.

Main Methods:

• Utilized statistical learning to derive quantile smoothing formulas.
• Integrated smoothed quantiles with bootstrap methodology.
• Applied the method to equivalence testing in mean and covariance structure analysis.

Main Results:

• Developed novel formulas for smoothing quantiles of chi-square type statistics.
• Demonstrated improved accuracy in quantile estimation compared to traditional methods.
• Successfully applied the method to real data for model misspecification quantification.

Conclusions:

• The proposed statistical learning approach enhances the reliability of chi-square type test statistics.
• Smoothed quantiles improve equivalence testing accuracy, particularly in complex analyses.
• The methodology offers potential for smoothing other statistical estimates and test statistics.