This looks at the issues of using the right models for linear regression. In this case I am using the type I ordinary least squares regression model for data where the x values are known and do not have errors.

A unified framework for quantifying holobiont-level redox resilience by integrating plant physiological traits, soil redox chemistry, and microbial community organization into a single interpretable index. The package implements the Redox Resilience Index (RRI), a weighted composite metric grounded in redox biogeochemistry, microbial ecology, and systems biology. Multivariate indicators from each domain are reduced to normalized one-dimensional latent scores using flexible dimension-reduction methods (e.g., PCA, factor analysis, nonlinear embeddings, or co-expression networks). Microbial resilience is represented as a blended latent component combining abundance or functional composition with network organization metrics derived from ecological graphs. RedoxRRI is designed for hypothesis-driven research rather than black-box prediction, enabling transparent comparison of alternative aggregation strategies, domain weights, and ecological contexts. The framework supports validation against ecosystem recovery, stress tolerance, and biogeochemical function across experimental and observational studies.

Demonstration the fundamental difference between Differential Gene Expression, DEG/limma, and eQTL/gene-specific genotype analysis.

Project 3

A five-slide presentation describing the next word prediction algorithm and the Shiny application developed for the Data Science Capstone (SwiftKey project).

Final Project