Polyunphased

Polyunphased extends Unphased to perform conditional polytomous logistic regression for genetic association analysis of polytomous phenotypes in nuclear families, providing tests robust to population stratification.

Key Features:

• Conditional polytomous logistic regression: Implements conditional polytomous logistic regression tailored for within-family association analysis of polytomous phenotypes.
• Robustness to population stratification: Provides within-family tests that are robust against confounding by population structure.
• Compatibility with Unphased: Retains core functionalities and most options available in the Unphased package.
• Handling missing parental genotypes: Includes options to handle missing parental genotypes while preserving robustness to population stratification.
• Maximum likelihood estimation: Uses maximum likelihood estimation for association parameters in the conditional logistic regression model.
• Parental-term parameter handling: Supports fixing parental-term association parameters at zero; independent estimation of these parameters showed convergence issues.
• Implementation language: Implemented in C++.
• Simulation-based validation: Simulation studies validated the statistical behavior of maximum likelihood estimates when parental-term association parameters are set to zero.

Scientific Applications:

• Family-based genetic association studies: Analysis of genotype-phenotype associations within nuclear families.
• Polytomous phenotype analysis: Association testing for diseases with multiple subtypes or multilevel categorical outcomes.
• Concurrent dichotomous-trait analysis: Analysis of two dichotomous phenotypes analyzed concurrently as a polytomous outcome.
• Robust association testing: Detection of genetic associations while controlling for population stratification and missing parental data.

Methodology:

Uses conditional polytomous logistic regression with maximum likelihood estimation and simulation studies to evaluate estimator behavior, with recommended practice of fixing parental-term association parameters at zero due to observed convergence issues when those parameters are estimated.