Given an PlasmoMAPI project with a map already assigned, plot the coverage (number of intersecting ellipses) of each hex. This plot acts as a diagnostic, as hexes with low coverage will conform less well to the assumptions of the permutation test (see details).
plot_coverage(proj, breaks = c(0, 10, 20, 30, 40, 50, 100, Inf))
| proj | object of class |
|---|---|
| breaks | the sequence of coverage breaks used. |
Good coverage is needed to ensure the validity of the statistical procedure. For each hex, the observed value is the mean of the normalised values of the edges that intersect it. Under the null model that these normalised values have no systematic bias, the distribution of this test statistic is approximately normal as a consequence of the central limit theorem. Hence, we can use a permutation test to characterise the mean and standard deviation of this null distribution, then we can quantify how extreme the observed value is in terms of its z-score. However, if coverage is too low then the null distribution will not be normally distributed, and hence the z-score will not be an accurate description of how extreme the observed data really is.