A simple powerful bivariate test for two sample location problems in experimental and observational studies

Area of Science:

• Biostatistics
• Medical Research Statistics

Background:

• Bivariate analysis is crucial in medical research for simultaneous testing of two equally important variables.
• Existing bivariate location tests have stringent assumptions (e.g., specific distributions, elliptical symmetry).
• A need exists for a powerful bivariate test that bypasses these restrictive assumptions.

Purpose of the Study:

• To propose a novel, powerful bivariate test statistic for comparing two populations.
• To develop a test that does not require assumptions of distribution, affine-invariance, or elliptical symmetry.

Main Methods:

• The bivariate problem was reduced to a univariate problem by summing or subtracting measurements.
• Monte Carlo simulation techniques were employed for comparison.
• The proposed test was compared against Hotelling's T(2), Rank test, Cramer test, and Mathur's test.

Main Results:

• The proposed test demonstrated superior performance compared to its competitors across most tested populations.
• It showed equivalent performance to the Rank test under specific distribution assumptions.
• The test proved more powerful than alternatives for various location shifts and directions.

Conclusions:

• The proposed bivariate location test is highly effective across diverse population distributions (normal, non-normal, skewed, symmetric, heavy-tailed, medium-tailed).
• Its robustness and power make it suitable for practical applications in medical research.
• The test is recommended due to its superior power over compared alternatives.