Hello everyone
I wonder if anyone can advise how I can set a minimum value for an interpolated surface created by kriging and cokriging? (I want to set a minimum predicted value of zero). I am using the Geostatistical Analyst Wizard to create my model specifications - but may be able to get the modification scripted if necessary.
Many thanks, Beth
Beth
In 10.1 we introduced some new transformations. If you pick LogEmpirical the predictions will be > zero.
-Steve
Many thanks for this Steve. I am working on 10.2.2 - in the wizard I see that I can choose log but can't see 'logempirical' please can you advise where this option is? Beth
Thank you very much Steve. Sorry, one more question - if you have time? I am using the predicted surfaces as a visual mapping tool to discuss with farmers possible management changes to their farms (I am modelling soil carbon). I assume using this log transformation will give skewed visuals - ie some parts of the range will be more emphasized (visually) than others? (I do understand that I can calculate the true values by back-transforming the predicted ones if I am just looking at the numbers).
In most cases there should not be a problem. You can, however, see how well or how badly the data are being transformed. On the NormalSoreTransformation page in the Wizard change to a Cumulative graph and you’ll see if there is a large deviation or not.
The Wizard automatically does the back-transform so you don’t have to do that yourself.
-Steve