One-dimensional data in spatstat

Question

I have data of one-dimensional points (locations in a straight line). I want to check if the cluster poisson process models or Cox models are a good fit for the data.

Since my data only has x-coordinate, I tried the linear network poisson process model. However, lppm supports only Poisson models.

The kppm methods, which have the cluster models, require two-dimensional data. So, I added a dummy y-column with zero values and range of (ymin=0,ymax=0.001) . ymax=0 returns errors during computation. Now, I am able to fit Matern cluster, etc.

My question is what would be the best way to handle one-dimensional data?

Is adding a dummy column with a non-zero range the only solution? Or is there some detail about point patterns or process models that I have missed? Suggestions for alternatives are also welcome.

Answer 1

First can I point out that other commentators have confused the Cox Point Process with the Cox proportional hazards model -- two completely different things, invented by the same person.

A Cox point process is a perfectly valid model for the data in this context.

Cox point process models for data on a linear network (like a street network) are not yet supported in the spatstat package. However you can fit such models using the minimum contrast fitting procedure. First compute the estimated 1-D pair correlation function or K-function from your data (using linearK or linearpcf ). Then use mincontrast to fit the desired model. You will have to write a small piece of code to compute the theoretical value of the K function or pcf for the desired model. See help(mincontrast) .