envelope-function and pre-simulated patterns: Number of patterns not recognized about spatstat HOT 5 CLOSED

Thanks - The number of simulations is specified by nsim, which defaults to 99, so this is not strictly a bug. However I guess what you're saying is that most users would expect 'envelope' to set nsim = length(simulate) when 'simulate' is a list of point patterns and nsim is not given. I guess we can do that. Thanks for pointing it out. A

from spatstat.

On second thoughts, this might be complicated. Sometimes a second set of simulations is required (eg global=TRUE and X is a point process). Then we want nsim + nsim2 = length(simulate). I'm not sure how to guess the user's intention easily. A

from spatstat.

In my usecase I want to compute envelopes from the pre-defined set of simulations. This always(?) excludes global =TRUE as

1. when X is 'ppm' the average can't be guaranteed to be correct for 'simulate'
2. when X is 'ppp' there is no model specified for deriving the average

from spatstat.

The following is possible, where the 'simulate' argument has to have 2 * nsim entries:
a <- rStrauss(100, 0.5, 0.05, nsim=38)
E <- envelope(cells, Gest, simulate=a, nsim=19, global=TRUE)

from spatstat.

imho it would be better behaviour if the nsim and nsim2 would adapt to length(a) . It would solve the original issue, and would not waste patterns as in e.g.

a <- rStrauss(100, 0.5, 0.05, nsim=38)
E <- envelope(cells, Gest, simulate=a, nsim=15, global=TRUE)

which assumes nsim2=nsim and leaves over the 38-15*2=8 patterns. Another reasonable scenario is the case "here's a bunch of patterns, compute the global average from 10 of them", i.e.

E <- envelope(cells, Gest, simulate=a, nsim2=10, global=TRUE)

which crashes as nsim=99 is fixed and nsim+nsim2 > length(a).

from spatstat.