Analysis of two-way layout of count data with negative binomial variation

Area of Science:

• Statistics
• Biostatistics
• Data Analysis

Background:

• Standard analysis of variance (ANOVA) methods have strict assumptions not always met in real-world data.
• Count data, especially from biological experiments, often violate normality assumptions.
• Alternative methods are needed for robust analysis of factorial experiments with non-normal count data.

Purpose of the Study:

• To propose a likelihood ratio statistic for analyzing factorial experiments with negative binomial count data.
• To provide a method for testing main effects and interactions when ANOVA assumptions are not met.
• To derive a statistic for comparing dispersion parameters across multiple groups.

Main Methods:

• Utilized the likelihood ratio statistic for hypothesis testing.
• Applied the method to two-way layout count data following negative binomial distributions.
• Derived a statistic for testing the equality of common dispersion parameters.

Main Results:

• The likelihood ratio statistic effectively tests main effects and interactions in negative binomial models.
• The derived statistic allows for comparison of dispersion parameters.
• The method demonstrated robustness in analyzing spatial and temporal bacterial count variations.

Conclusions:

• The proposed likelihood ratio statistic offers a valid alternative for analyzing factorial experiments with count data.
• This method is particularly useful when data deviates from normal distribution assumptions.
• The approach provides valuable insights into spatial and temporal variations in biological count data.