RobPower

RobPower computes analytical expressions for statistical power and type 1 error rates in genome-wide association studies (GWAS) under mixed-model frameworks.

Key Features:

• Algebraic Formalization: Formalizes statistical power and type 1 error rate calculations for classical GWAS methods including simple regression, GRAMMAR, FASTA, and the transmission/disequilibrium test for quantitative traits.
• Analytical Formulae: Derives analytical formulae using matrix algebra to compute the first and second moments of test statistics under a true mixed model that includes polygenic effects and SNP effects.
• Customizable Parameters: Accepts relationship matrices between phenotyped and genotyped individuals and heritability levels as inputs to tailor type 1 error rate and power computations to specific designs.
• Method Comparison and Evaluation: Evaluates method performance, reporting that regression type 1 error rate increases with relationship variability and heritability, GRAMMAR reduces this inflation, and FASTA and quantitative transmission/disequilibrium tests remain unaffected by those factors.
• Generalized Results: Generalizes simulation study results from the literature to provide broader applicability of method performance across different study designs.

Scientific Applications:

• Type 1 error estimation: Computes type 1 error rates for SNP–phenotype association tests in GWAS and association mapping studies.
• Power calculation and sample design: Calculates statistical power to inform sample size and study design decisions for detecting quantitative trait loci.
• Method selection: Compares regression, GRAMMAR, FASTA, and transmission/disequilibrium approaches to guide selection of statistical methods based on relationship structure and heritability.

Methodology:

Derives analytical expressions using matrix algebra to compute first and second moments of test statistics under a true mixed model including polygenic and SNP effects; incorporates relationship matrices and heritability estimates into these calculations and computes thresholds for type 1 error rates and statistical power.