0

Following the steps in the answer given here, I've fit Thomas cluster model parameters kappa and sigma to the pooled Kinhom functions of a set mppm intensity predictions. By "transferring" the inhomogeneity to the mu parameter (as a function converted to a pixel image) and solving for a scale factor so that that kappa*sum(scaled_mu) equals lambda, I have a complete set of parameters for a single model derived from the pooled data. Is there a way to construct a kppm object from this parameter set in order to begin evaluating model performance? Or do I need to rely on simulated realizations of the model parameters for comparison with my actual data?

(A possibly relevant extra piece of information is that the point patterns to which I'm fitting the model are each observed on different polygonal windows.)

Thank you!

SGJ
  • 1
  • 1

1 Answers1

1

No, this won't work.

You could create a fake kppm object (i.e. an object of class "kppm" which has the parameters you've obtained, except that it was not generated by fitting the model to a particular point pattern dataset using kppm()). However the internal data in this object is not self-consistent; methods for class "kppm", such as predict.kppm, simulate.kppm, will not work correctly.

What you're asking for is effectively the counterpart of mppm for cluster process models. This has not yet been developed. It is on the "to do" list.

Adrian Baddeley
  • 2,534
  • 1
  • 5
  • 8