Hello, I am using the the Logistic Regression in LCM - TerrSet and I wonder, for categorical variables, if I should use 'dummy' rasters (i.e., with 0 and 1 for the landscape features of interest) or if I should first transform the categorical rasters using the Evidence Likelihood method in LCM and then use the transformed raster as a variable in the Logistic Regression?
Which one would be the most theoretically sound approach to take? I am worried that the Evidence Likelihood transformation somehow compromises the interpretation of the parameter estimates associated with a particular variable.