Dear Karolina,

Thanks for your interest in MGET. You can see an example workflow here and also in the MGET paper. The tools have evolved somewhat since these were created—the names and parameters are somewhat different—but you should get the overall idea. Also, these examples are with GAMs but the random forest tools are very similar.

The formula specifies the fields of your table that are the response variable and predictor variables in syntax that is traditional for the R programming language. In short:

ResponseField ~ PredictorField1 + PredictorField2 + ... + PredictorFieldN

Most likely you are fitting a classification model. In that case, you need to specify the response field enclosed in factor(), like this:

factor(ResponseField) ~ PredictorField1 + PredictorField2 + ... + PredictorFieldN

For example, if you were classifying benthic habitat type from physiographic variables, you might say:

factor(HabitatType) ~ Depth + Slope + Rugosity

HabitatType should be an integer or text field of codes. Depth, Slope, and Rugosity would be continuous variables in floating point fields. If you have a categorical predictor variable (an integer or text code), it is necessary to enclose it in factor() as you do when you have a categorical response variable:

factor(HabitatType) ~ Depth + Slope + Rugosity + factor(SubstrateType)

To learn more, be sure you click the Show Help>> button in ArcGIS. Each parameter of the tool has help text that describe what it is for.

Let me know if you have any more questions. Good luck!

Jason