A semiparametric quantile regression rank score test for zero-inflated data.

Area of Science:

• Statistics
• Biostatistics
• Data Science

Background:

• Zero-inflated data are prevalent across diverse scientific fields, posing challenges for standard statistical models.
• Traditional methods like Zero-Inflated Poisson and Negative Binomial models often rely on restrictive parametric assumptions.
• These assumptions may not hold in real-world scenarios, limiting their applicability and accuracy.

Purpose of the Study:

• To introduce a novel semi-parametric statistical test, the zero-inflated quantile single-index rank-score-based test (ZIQ-SIR).
• To address the limitations of existing methods in analyzing zero-inflated data with potential nonlinear covariate relationships.
• To provide a flexible and robust approach for association detection in complex datasets.

Main Methods:

• Development of the zero-inflated quantile single-index rank-score-based test (ZIQ-SIR).
• Utilizing a rank-score-based approach to accommodate semi-parametric modeling and avoid strong distributional assumptions.
• Evaluating performance through extensive simulations and application to a real-world microbiome dataset.

Main Results:

• ZIQ-SIR demonstrated superior statistical power and improved Type I error control compared to existing methods in simulations.
• The method effectively handles both zero-inflation and overdispersion commonly found in count data.
• Application to the Columbian Gut study's microbiome data revealed more significant associations than alternative approaches.

Conclusions:

• ZIQ-SIR offers a flexible and robust semi-parametric alternative for analyzing zero-inflated data, particularly with nonlinear relationships.
• The proposed method outperforms traditional parametric models in terms of power and error control.
• ZIQ-SIR provides valuable insights into complex biological data, as evidenced by its application to microbiome abundance data.