linear regression of multiple rasters

[ShawnONeil]

I'm wondering if anyone can help me.  I would like to perform regression analysis on raster data where the model is run on a cell-by-cell basis.  I have multiple raster layers created in ArcGIS representing various habitat attributes for a wetland study area.  We are trying to model these layers as explanatory variables for home ranges which are represented by a kernel density layer (utilization distribution or UD).  The UD is a continuous surface representing the probability of an animal using the space at each particular pixel.  I need to relate this space use to the habitat pixels at the same locations.  I've tried aligning and clipping each layer to the UD and exporting the result as a text file for analysis in R, but even when perfectly aligned the number of columns and rows in the resulting grid is not quite the same as the UD I've clipped it to.  Therefore, I'm worried that the value in one table won't correspond to the correct value in the next table.  Does anyone have any ideas, or know someone who might be an expert?  I've tried to explain this the best I can, please let me know if I need to clarify.

Thanks!   

Shawn O.

[EricRice]

Hi Shawn,

When you clip the rasters to your UD raster, make sure a couple other parameters are set.  You can find these in the environment settings.

1. Cellsize - set this equal to your UD raster.
2. Extent - set this equal to your UD raster.
3. Snap Raster - set this equal to your UD raster.  This is a very important setting as it will guarantee cell alignment by snapping the output to the UD raster.

Regards,
Eric

[ShawnONeil]

Thanks Eric,

I had equal cell sizes with the extent set and snap raster settings set to the UD as well, and was still getting some edge effects where there would be an occasional mismatch in pixels between layers.  I think I've discovered the solution, which was to change environment settings so that the extent was "intersection of inputs" and XY tolerance was .0e-013.  This gives me the same boundary for each home range and the same # of row & columns in each output file.  I still need to test it on the rest of the data, but I'm confident it will stay consistent.

Shawn