Fig. 4.

Graphical hypothesis testing: Global envelope test, ‘extreme rank’ type, two-sided at 95% significance level using 2500 simulations of the null hypothesis ($H_0$: no spatiotemporal clustering nor inhibition, i.e. $\tau =1$), $p$-value $\in [0,0\text{⋅}022]$. Note there is a region where $\hat{\tau }$ just exits the global envelope lower bound (suggesting inhibition at $\sim$100 m) as well as the obvious departures above the upper bound (suggesting clustering at close distances and $\sim$190 m). We are confident that we are simulating $H_0$ well because the median simulation stays close to $\tau =1$ throughout. Distance band set $≔\left\{[0,10),[0,12),[0,14),\dots ,[0,50),[2,52),[4,54),\dots ,[170,220)\text{m}\right\}$.