Log-Transform the data

[GladstoneAlexandre]

Dear,

I have a question about the processing of Log-Transform the data...

When I use the method to Log-Transform the data, the results presented in the variogram analysis are still presented in log. How I do for turning the value of partial sill, for example, for my original unit?  I tried to do the inverse of the log but it did not work.

Thank you in advance.

[EricKrause]

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.

[Elijah]

Hi Eric,

I was searching to know how exactly the (ArcGIS) software back-transforms log-normalized data into the original scale. I am particularly interested in how this is handled in EBK, in case it's handled differently in other models. Please, can you explain to the extent possible?

Thanks

Elijah

[GladstoneAlexandre]

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 will be in the original untransformed units.

Dear Eric,

Thanks for the reply. But do I need to know the actual value of the nugget effect and partial sill to calculate the correlation of my spatial analysis (C0 / (C0/(C0 + C1)) and display the values �??�??in the work that I am developing.

Thanks again.

[EricKrause · Accepted solution]

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.

[GladstoneAlexandre]

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.

Thanks for the explanations Eric. I thought there were some way to do the reverse of the results and get the values �??�??in real units.

Until next time.