Geostatistics analysis for non-geographic data in ArcGIS

[ChaoWang5]

Hi All,

I am a biologist. Currently, I am trying to use a geostatistics algorithm (kriging algorithm) to analyze my experimental data - to illustrate how protein behaves and functions in the Cellular 'geographic map'. So I don't have latitude and longitude values, instead, I have a set of (X,Y,Z) data indicating different function of a protein at different cellular locations. The scale of the (X,Y) is quite different from latitude and longitude. The question I am asking is - can I analyze non-latitude and longitude value in the Arcmap software? How can I do that?

Thanks,

Chao.

[SepheFox]

I Don't see why not. Arcgis happily accepts custom coordinates. Have you tried importing your data yet, and creating features with z values from your table? I'm curious to know the result.

[SepheFox]

IF your not familiar with the process, the following article will help. The last section at the bottom explains how to convert a table with z values. I'm Not sure what you would do for picking a coordinate system. Depending on your scale you could fake it in meters.

[ChaoWang5]

Thanks for your reply, Sephe. Where is the 'following article' you mentioned in your post? I can't find it.

I tried my data in ArcGis. However, I found my X and Y scale are very different. The X-scale is from 0 to 1278. But the Y-scale is from 0 to 1. Thus the map in ArcGis is just a very narrow bar. Can I modify the X and Y scale in ArcGis? This is the first time I am using ArcGis......I have no idea how to modify the X and Y coordinates.

In fact, I used another software called 'Originpro' to performed Kriging algorithm for my data. The result is attached below. However, this software does not output the semivariogram or 'confidence map', so I want to use ArcGis to produce semivariogram or prediction standard error.

Thanks

Chao.

[SepheFox · Accepted solution]

Sorry Chao, I replied on my phone last time, and I guess the link didn't get attached like I thought. Anyway, it sounds like you know how to convert the x/y values, no problem. If you're not concerned with preserving the actual locations of the points, then I would scale the y values by multiplying by some factor. In your case probably 1000.

[ChaoWang5]

Thanks. I think that would work.