# Inverse distance weighted interpolation for categorical and/or binary datasets

I want to perform an `Inverse distance weighted (IDW) interpolation` to a series of points. However my data set can be `categorical` or `binary` and just like the `IDW` I want to interpolate the new values (categorical or binary) based on the highest frequency of closest points values surrounding the prediction location:

A if the majority of the closest points are A

C if the majority of the closest points are C

...

In addition I want to avoid using close points that are behind a barrier, that can be a wall or just an elevation.

This is how the problem/dataset looks like:

The grey points is the point I want to interpolate and the points behind the barriers can't be used in the process. The output of the interpolation process would be A or B, in this case probably A, since the closest points are all A.

What I am looking for is some kind of work-flow that would allow me to do this in ArcGIS, if possible I would prefer R (if anyone has any experience in doing this kind of processes in R). This is just the example how my problem looks like, I still have to implement it to a bigger extent and to many layers of points. Bear in mind that I have little experience with interpolation.

• The barrier definition is a bit vague. At least two more red points on the right are shadowed by the barrier from the grey point. They're not exactly 'behind' but the grey point can't see them. Or do those grey parallel lines indicate some kind of directional effect of the barriers? Jan 24, 2013 at 18:45
• @Spacedman yes indeed there are some other red points that are not visible from the grey point, therefore can't be used. The drawing was just to give an idea of what I am looking for. The grey parallel lines indicate areas that can't be used in the interpolation, because they are not visible from the grey point (e.g. behind a barrier or on top of a barrier, the barrier can be for example a mountain) Jan 25, 2013 at 9:43
• So to be clear if you draw a line from the gray point to every other point, any line that intersects a barrier feature is not included at all in the analysis for that gray point? Sometimes barrier features just mean the distance measure has to go around it, so that points behind barriers are further than straight line distances, but not effectively non-existent. Jan 25, 2013 at 10:39
• A quick google search brings up this paper: sciencedirect.com/science/article/pii/S0016706107001218 - seems to use geostatistical methods and clustering rather than IDW. Jan 25, 2013 at 10:42