countfitteR

countfitteR performs Bayesian selection of count-distribution models to improve statistical analysis of discrete biological count data such as DNA double-strand break foci and molecular markers including phosphorylated histone H2AX and p53 binding protein.

Key Features:

• Bayesian Model Selection: Uses a Bayesian approach for objective comparison of competing count-distribution models.
• Supported Distributions: Compares Poisson, negative binomial (NB), zero-inflated Poisson, and zero-inflated negative binomial distributions.
• Zero-inflation and Overdispersion Handling: Accounts explicitly for zero-inflation and overdispersion in count data.
• Estimation and Uncertainty: Estimates mean values and confidence intervals under the selected distributional model.
• Comparative Performance: Reports statistical performance versus traditional two-step procedures, with reported overall power up to 98%.
• Input Data Types: Applicable to discrete count data from imaging quantifying DNA double-strand breaks and molecular pharmacological markers such as phosphorylated histone H2AX and p53 binding protein.

Scientific Applications:

• DNA damage quantification: Quantification of DNA double-strand breaks measured as discrete foci in imaging datasets.
• Cancer and aging research: Analysis of count-based biomarkers in cancer and aging studies, including assessment of drug efficacy.
• Molecular pharmacology: Statistical analysis of counts for phosphorylated histone H2AX and p53 binding protein as molecular markers.
• Non-Poisson count analysis: Analysis of datasets exhibiting zero-inflation, overdispersion, or other departures from Poisson assumptions.

Methodology:

Implements Bayesian model selection to compare Poisson, negative binomial, zero-inflated Poisson, and zero-inflated negative binomial models and to estimate mean values and confidence intervals while addressing zero-inflation and overdispersion.