New statistical method for analyzing sparse, overdispersed count data introduced

Researchers have introduced a new statistical method called Dirichlet-multinomial (DM) deviance residualization designed for sparse, jointly overdispersed count matrices. This approach is particularly relevant for biochemical assays that generate such data. The DM null model treats count vectors as fixed-total compositions, with a single concentration parameter governing overdispersion, and is derived by conditioning independent negative-binomial feature counts on the observed sample total. This method preserves sparsity, offers constant-time evaluation per nonzero entry, and shrinks residuals based on the tolerated overdispersion, with the multinomial residual being recovered as the dispersion parameter approaches infinity. AI

RANK_REASON The cluster contains a new academic paper detailing a statistical methodology. [lever_c_demoted from research: ic=1 ai=0.4]

Read on arXiv stat.ML →

• Akshay Balsubramani
• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Dirichlet--multinomial
• Gotit.pub
• Hugging Face
• Influence Flower
• ScienceCast

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →