Latest Contributions by EricKrause

‎06-27-2013

I wouldn't go so far as to say that they "can't" be tested. However, using a correlation coefficient that assumes independent samples is not a valid way of testing it. You should just acknowledge that you are ignoring spatial correlation and calculating the correlations for exploratory purposes. The reason I wouldn't say that they "can't" be tested is that I've read a few papers about testing correlations between two spatially-correlated rasters (such as this one), but we don't have pre-built tools to do it in ArcGIS.

‎06-25-2013

This is a strange error to get from GWR if you only have one explanatory variable. Is it possible to send your data to ekrause@esri.com so I can take a look?

‎06-25-2013

I think it has to do with the large areas of zero snowfall. GWR needs local variability to compute estimates, and large areas of constant value will cause the tool to give that error. As for what you can do, I honestly don't know. Zero-inflation is a huge problem for statistical models. You might find some success with performing GWR (or cokriging) only in the areas where you have snowfall, then trying to fit a different model to the areas with no snowfall.

‎06-25-2013

You can input all of your density rasters into Band Collection Statistics to get a correlation matrix between all pairs of rasters. However, keep in mind that Pearson correlation coefficients assume independent samples, and your density maps will have spatial autocorrelation. Be careful before using these correlations to calculate p-values.

‎06-21-2013

To use Geographically Weighted Regression, you would need to convert the raster to points, then perform a spatial join to get them all in the same feature class.

‎06-13-2013

Yeah, this sounds like a case where cokriging might be effective. Cokriging is most effective when the secondary variable is much more finely sampled than the primary variable, and it sounds like that is what you have. If you find that cokriging is not effectively using the precipitation, you may want to try Geographically Weighted Regression.

‎06-11-2013

Yes, both of those graphics are saying the same thing using different notation. The only difference is that the first graphic is giving the formula for using exactly 7 neighbors (and it uses "n" to indicate the power, which is a bit unusual). The second graphic is the general formula for any number of neighbors.

‎05-28-2013

If this is a functionality you would like, you should suggest it on the ideas site.

‎05-28-2013

This is a tough call. Both rules are important, but I tend to give more preference to crossvalidation. However, since you have one model with a perfect visual fit and another model with perfect crossvalidation, I think you can find a third model that has both. Try using Spherical and K-Bessel semivariograms.

‎05-14-2013

You need to understand a few things. By applying a log transformation in kriging, you're saying that the logarithm of the data values follows a Gaussian spatial process whose covariance structure can be modeled with a semivariogram. This doesn't necessarily imply that the untransformed values can be correctly modeled with a semivariogram. In this case, it only really makes sense to talk about correlations, sills, nuggets, etc of the log-transformed data. If you want to try to make these statements about the untransformed data, you need to try to model the data without a transformation. But again, this may not be possible, depending on your data.

‎05-14-2013

I don't know of any direct way to compute the sill of the untransformed data from the sill of transformed data (other than just not transforming and calculating it). But I'm not clear on why you want to do this. The software models the semivariogram in the transformed space, but it does the back transformation automatically. The resulting surface will be in the original untransformed units.

‎05-06-2013

You're right to avoid kriging with only 14 points that represent percents. Unfortunately, you probably won't be able to fit a reliable/accurate model with so few points no matter which method you use. If you absolutely have to interpolate them, you should try as many interpolation methods as you can and hope one of them stands out from the rest. You may want to try Natural Neighbors from Spatial Analyst as well.

‎04-25-2013

The epsilon is the residual (or error) in the prediction. Recall that OLS does not make perfect predictions, and epsilon is the difference between the true value and the predicted value.

‎04-25-2013

Yes, B 0 is the intercept term.

‎04-24-2013

Areal Interpolation will often predict outside of the data range. This means that sometimes you can get negative predictions for a phenomenon that can't be negative. Unfortunately, there isn't much you can do to prevent this. Does your data represent counts? If it does, you should use Event (overdispersed Poisson) areal interpolation. This type of areal interpolation will never predict negative values.

Online Status	Offline
Date Last Visited	‎02-25-2026 06:39 PM

My Ideas

Latest Contributions by EricKrause

Re: Spatial correlation between layers

Re: Interpolating snowfall point data using a precipitation raster as a guide

Re: Interpolating snowfall point data using a precipitation raster as a guide

Re: Spatial correlation between layers

Re: Interpolating snowfall point data using a precipitation raster as a guide

Re: Interpolating snowfall point data using a precipitation raster as a guide

Re: Which equation is used for the IDW interpolation?

Re: MapTips (predicted value at cursor) rounding

Re: Which model is better in GA?

Re: Log-Transform the data

Re: Log-Transform the data

Re: Trying to find best interpolation method for small dataset

Re: OLS results interpretation

Re: OLS results interpretation

Re: Negative values after areal interpolation

Re: Why exporting EBK 3D to voxel layer change org...

Re: Access geostatistical layer created using ArcP...

Re: Why exporting EBK 3D to voxel layer change org...

Re: K-Bessel Semivariogram Equation

Re: Kriging Model Types

R-ArcGIS