Hello

I am fairly new to GIS and i am trying to run a regression analysis on two different point data drawings. Can anybody point me in the right direction?

I have two seperate drawings which have X, Y & Z points in the same filed however the points are not exactly on top of each other. I want to run a regression analysis between the two data sets to see the correlation between the two within a 5m (approx) spatial area.

I have made the points into a surface however i am lost at my next move.

Any help would be appreciated

Pete.M

Manifold in itself doesnt provide regression analysis functionality. You will need to get for example GeoDA or even use R to do the regression analysis per se.

As to you transforming your pointset to a surface, I dont know how good an idea that is? The interpolation algorithms Manifold offers are a bit of a black box, and you introduce a level of uncertainty into your data by the transformation. I would simply create 5m buffers of the one pointset, and then use a spatial overlay to transfer the values to the other pointset.

Just interested, what would be your dependant variable? do you have multiple candidate independant variables?

Check out the Manifold Wiki with SQL and scripting examples at http://www.manipedia.eu/

Spatial Knowledge, my personal blog.

Thanks for the reply petzlux. The only reason why i was trying to create a surface is that the data capture interval is not the same. One data set has been captured every second on a of 15m swath over 100 hectares of land. The second dataset has been captured every 3 seconds on a 2m swath .

Pete.M

If the two datasets represents spatial samplings of continous phenomena, then why dont you transform both to a surface?

I am no expert in the analysis of such processes, but a good reference would be the following website:

http://www.spatialanalysisonline.com/output/

Check out the Manifold Wiki with SQL and scripting examples at http://www.manipedia.eu/

Spatial Knowledge, my personal blog.

Although Manifold does not directly furnish regression tools, with the magic of Microsoft automation techniques, you can directly connect to Excel's object model and do what you are talking about. You don't need to exit Manifold, you can script an ActiveX connection to Excel and export your array(s) to Excel and bring the results back into a Manifold table and then into a drawing, etc.

You need to be able to write a script to do this. Art Lembo's GisAdvisor.com website has an excellent tutorial for sale, called Manifold Scripting in version 6.5, I think, that has an good example on how to do this. In fact, he uses Excel in the example to run a correlation analysis, a close cousin of your regression analysis, so it should be fairly easily to modify for your needs. It's really not that hard, of course, after you see it done.

There are many examples on the web showing how to automate Excel from within other programs, if you don't want to purchase Art's tutorial. I recommend the tutorial, it's cheap for the value.

It seems that Manifold is not the environment for this analysis. Do you mean that you have pairs os points 5m apart each other? your variable of interest is Z values?

I don`t know if i understand it really, but why you don`t make a voronoi of your 2 Point-Drawing (or work with buffers & bounded boxes), so you get an area drawing. After this you can do a topology overlay and you have both values in one drawing ?

http://www.thegisservicesector.de