[R-sig-Geo] uniform weighted (non-Gaussian) kernel in density.ppp

[Adrian Baddeley]

Jan Quets <Jan.Quets at ua.ac.be> heeft geschreven:

> is it possible to define a circular kernel within the function density.ppp(),
> that assigns uniform weights to every point inside the kernel.

I will add a suitable function to the next version of spatstat. 

This is not feasible using the current algorithm in density.ppp, which is based on the Fast Fourier Transform. To use the FFT we need a kernel which has finite bandwidth in the frequency domain. The uniform kernel on a disc has a sharp cliff at the circular boundary, so it has infinite frequency.

To compute the convolution with a uniform kernel requires a separate, slower algorithm (identical to the 'scan statistic' and implemented in non-R software like SatSCAN). This algorithm is already implemented in C inside spatstat, and requires only an R interface, which I will add in the next version of spatstat.


Adrian Baddeley