Testing Differences Between Pathogen Compositions with Small Samples and Sparse Data

The structure of pathogen populations is an important driver of epidemics affecting crops and natural plant communities. Comparing the composition of two pathogen populations consisting of assemblages of genotypes or phenotypes is a crucial, recurrent question encountered in many studies in plant disease epidemiology. Determining whether there is a significant difference between two sets of proportions is also a generic question for numerous biological fields. When samples are small and data are sparse, it is not straightforward to provide an accurate answer to this simple question because routine statistical tests may not be exactly calibrated. To tackle this issue, we built a computationally intensive testing procedure, the generalized Monte Carlo plug-in test with calibration test, which is implemented in an R package available at https://doi.org/10.5281/zenodo.635791. A simulation study was carried out to assess the performance of the proposed methodology and to make a comparison with standard statistical tests. This study allows us to give advice on how to apply the proposed method, depending on the sample sizes. The proposed methodology was then applied to real datasets and the results of the analyses were discussed from an epidemiological perspective. The applications to real data sets deal with three topics in plant pathology: the reproduction of Magnaporthe oryzae, the spatial structure of Pseudonzonas syringae, and the temporal recurrence of Puccinia triticina.