[R-sig-Geo] `kest` radius limits?

[Nick Eubank]

Thanks Adrian. I think my confusion is that there are lots of points that
are more than r units away.

Sweden is basically a tall, thin rectangle. It's ~260km wide at its
narrowest point, but it's 1400km tall. And my point density is pretty high,
so for any point, there are lots of other points at more than 160km (just
not in all directions). In other words, the maximum distance from any point
to the boundary is NOT 160km. Thus my confusion.

Also, I've been overriding the `xlim` too -- the x-axis of the plot window
runs all the way to the value I want (300km), and the null poisson model
runs to 300km, it's just the estimates of k(r) that stop.

I don't think I can send pics to the r-geos list, but I'll email the image
to you.

(Ordered your book, btw)



On Wed, Jul 27, 2016 at 3:36 AM Adrian Baddeley <
adrian.baddeley at curtin.edu.au> wrote:

> Nick Eubank writes:
>
> > I'm running kest on points in Sweden (which is about 200km wide at its
> > narrowest) with 'border' correction and lots and lots of points.
>
> > However, I can't get estimates of K(r) for radii beyond ~160km. I bumped
> > `rmax`, and if I plot my results, the null model line extends  to the
> > `rmax` value, but the K(r) estimates just stop at ~160km.
>
> For the border correction, the estimate of K(r) becomes NA (and is not
> plotted) as soon as there are no data points that lie more than 'r' units
> away from the boundary. If the maximum distance from any data point to the
> boundary of Sweden is 160 km, then this is the behaviour I would expect.
>
> All explained in chapter 7 of the spatstat book.
>
> If you print a summary function object, the output shows the "recommended
> range" of distance values and the full "available range". By default, the
> plots of summary functions use the "recommended range" of distance values
> rather than the full "available range". To override this, when you plot the
> summary function, use 'xlim' to control the range of 'r' values. (xlim
> should be an argument in the 'plot' call).
>
> Adrian Baddeley
>
>
>

	[[alternative HTML version deleted]]