CODEWITHSUNDEEP
rspatstat
E. G. Mitchell
E. G. Mitchell

Reputation: 11

R: spatstat: Residual analysis for Thomas cluster processes

I wish to test the fit of a homogeneous Thomas cluster process, ideally using residual plots like diagnose.ppm and qqplot.ppm. Since the most straightforward way to fit a Thomas cluster process is using kppm, I can't use the above functions. Can anyone suggest ways to do this? e.g. transforming kppm to ppm or using ppm to fit a Thomas cluster process, Many thanks

Upvotes: 0

Views: 367

Answers (1)

Ege Rubak
Ege Rubak

Reputation: 4507

Just to possibly close up this old question:

There is no way to do this. It is not a implementation issue, but due to the fact that the ppm class (to which the functions diagnose.ppm and qqplot.ppm apply) represents fitted Gibbs point process models, which are fundamentally different than the fitted Thomas cluster process model (of class kppm).

Typically, goodness-of-fit for a Thomas process is judged using something like simulation envelopes based on summary statistics, but one should be careful not to use a summary statistic closely related to the one used to fit the model via minimum contrast (if that fitting method was used).

Upvotes: 0

Related Questions