regression advice needed

[ChristopherJursa]

Hello,

I have a series of scattered points on an area of interest with an attribute measuring the amounts of deposit of a mineral.  I would like to interpolate a grid using the points with their amounts and find hypothesized amounts at other surrounding locations.

I was looking at the Spatial Analyst tools and noticed Kriging.  However, I do not have elevations (Z values).  Can anyone recommend a tool in Spatial Analyst or ArcGIS for that matter? 

Thanks in advance!

[JeffreyEvans]

Applying a kriging model without first understanding the underlying spatial structure would be very questionable. Normally this done through an experimental semivariogram. You can then fit an empirical variogram and apply to a kriging model. It is imperative that you have a spatial signal and not a pure random effect (pure nugget). If you have a pure nugget effect kriging decomposes into a linear regression providing invalid results and defeating the entire point. In ArcMap, to perform this type of analysis you need Geostatistical Analyst. I would explore other interpolation techniques such as spline or IDW. If you have access to 3D analyst the natural neighbor (NN) interpolation provides very nice results. NN uses a tessellation approach making it very desirable if your underlying spatial structure is unknown.

[YaoYan]

It seems the interpolation is the right solution for the question since he asked to "predict" the mineral amount at other locations (that matches the original idea that enigeerer Krige designed the Kriging method).

As to regression, basically it CAN not predict at other locations unless there are addition infomation avaiable for the study area.