For more information, check the published paper by Naimi and Araujo (2016) in the journal of Ecography. This document provides a very quick demonstration on the package followed by some examples, that would be helpful to get a guick start with the package. # in the above example, the performance statistics were calculated based on the training dataset # as you can see, a report is generates shows how many percent of models were successful, and # model performance (per species), using training test dataset: # model run success percentage (per species) : # names of modelling methods : glm, gam, brt # in the following example, we use 3 different methods to fit the models. It is a better idea to have an independent dataset #(the data that were used to fit the mdoel). However, for most of cases, there is no such #(if so, we would specify in the test argument of sdmData). ![]() ![]() ![]() # data available, therefore, we can split the dataset as an alternative solution. P2 <- predict(m2, newdata=preds, filename= 'p2.img') # model Mean performance (per species), using test dataset (generated using partitioning): # toral number of replicates per model : 2 (per species) # thods (data partitioning) : subsampling # names of modelling methods : rf, cart, fda, mars, svm # Here we are going to fit 5 models and evaluate them through 2 runs of subsampling, each draw 30 percent # (i.e., subsampling, cross-validation, bootsrapping) There are also several methods to do that # be one time or several times (several replicates).
0 Comments
Leave a Reply. |