Correlation and Reclassifying Issues

[ColbyJackson]

Hello,

I'm having a few problems calculating the correlation of two grids within ArcInfo Workstation. When I preform the correlation I keep getting a Divide by Zero Error.

These are the steps I've taken so you can fully understand
1. I Convert two layers (Plant Types and Soil Type) to a Grid format using ArcGIS 9.3 Spatial Analyst
2. I reclassify one of the layers so I'm isolating one particular type of plant (The other values are given Nodata)
3. The reclassified layer is now used as a mask for the Extract by Mask function on the soil types layer. This will produce a soils grid of just the area occupied by that particular plant
4. The particular plant grid and the soils grid are now correlated within the Grid function of ArcInfo but everytime there is an error that states divide by zero is not possible.


If I reclassify the values as a zero instead of NoData in Step 2 the correlation will work but this doesn't focus the correlation on just the areas occupied by the plants but rather a neighborhood around them.
Also all cell sizes and extents were consistent throughout the analysis (set through the environments)

Any Ideas as to why this isn't working?

[ColbyJackson]

Okay This is what I'm trying to do.

I have a polygon shape file of both plant types and soil type. Within the plants shape file there are four different types of plants. I want to isolate each plant type to preform four separate correlations against the soil type it is growing in. The correlation method I am using is within Arc Info Workstation using the grid function.

I convert the plant type shape file to a raster (Spatial Analyst tools) and reclassify it four times to isolate each specific plant type. For example Queens Anne Lace is given the value of 1 while everything else is given a NoData (This is repeated for each plant type). This new plant classified raster is now use as a mask for the Extract by mask function on the Soils Raster that was created. It will allow only the areas where the plant is found to be correlated for soil types.

Every time I input the reclassified plant type and extracted soils grids into Arc Info to receive a correlation coefficient I receive the divide by zero error.

Also what do you mean by "It sounds like you are reducing your grids to those having a single (non-Nodata) value. Because that value exhibits no variation, the (Spearman) correlation is undefined?"

[ColbyJackson]

Yes I am trying to calculate a Pearson Correlation. I want to see if their is a spatial relationship with the plant types growing in a specific soil. If it can't be done this way, what would you suggest I do?