I'm working on a problem to understand the relationship between two point-based datasets with different sampling patterns.

We surveyed random subset of properties in a neighborhood for the presence of object X, and then used aerial imagery to tag all the objects of type Y we could see. So objects of type X have both a binary value (present or absent from a property) and a magnitude value (how many); type X have lat/lons at the property scale.

Objects of type Y have lat/lons at an scale equal to the resolution of the imagery, which is like 30cm - 1m, so much finer than property scale. The aerial imagery also covers the whole neighborhood, e.g. all properties, not a subset of properties.

I'm trying to determine if there's any relationship between the 1/0 value or magnitude of type X based on the (number within a radius? clustering?) of visible objects Y, but I don't have much of a background in spatial statistics.

I've tried adapting methods like KCross.Inhom & creating my own inhomogenous poisson process etc, but I don't really know if this is a valid approach given the different spatial scales of the data and the results are difficult to interpret.

I've done an adaptive density interpolation over all objects of Type Y but don't really know how to compare if the value of type X has any relationship to the density (positive or negative).

Essentially, I want to do a montecarlo simulation giving all houses in the neighborhood a 1/0 value for object X and noting their proximity to (the denisty? clustering?) of object Y and then see if the pattern I actually observed is random or significant.

Any thoughts?

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.