An effective sample size for predicting plant disease incidence in a spatial hierarchy

Area of Science:

• Agricultural Science
• Plant Pathology
• Biostatistics

Background:

• Predicting aggregated or heterogeneous disease incidence requires understanding disease distribution at different scales.
• The beta-binomial distribution is a common tool for modeling such data, but can be complex to apply.
• Existing methods may not be easily rearranged for predicting incidence at lower scales from higher scales.

Purpose of the Study:

• To derive a simple approximation for the beta-binomial-based function for disease incidence prediction.
• To introduce a new parameter (v) that relates to disease heterogeneity and sample size.
• To enable practical group sampling strategies for heterogeneous disease data.

Main Methods:

• Derived a novel approximation based on the binomial distribution, replacing the number of individuals per sampling unit (n) with a parameter (v).
• Determined the value of v iteratively by equating the zero-term probabilities of the binomial and beta-binomial distributions.
• Validated the approximation using field data of Eutypa dieback in grapes and simulated data.

Main Results:

• The approximation function successfully predicted disease incidence in aggregated and heterogeneous datasets.
• The parameter v, inversely related to disease heterogeneity, provides insights into data structure.
• The approximation allows for prediction of incidence at lower scales from higher scales, unlike the original beta-binomial function.

Conclusions:

• The derived approximation offers a simpler and more practical method for analyzing heterogeneous disease incidence data.
• This approach enhances the utility of group sampling in agricultural and epidemiological studies.
• The method is robust and validated for real-world disease monitoring and analysis.