MaxNet: Model module: MaxNet
In zoonproject/modules: Species Distribution Modelling Code Modules for the ZOON Package

Description Usage Arguments Details Version Date submitted Data type Author(s) See Also

Model module to fit MaxEnt models using the maxnet package

1 2	MaxNet(.df, features = "default", regmult = 1, clamp_predictions = TRUE, prediction_type = "logistic")

`.df`	Internal parameter, do not use in the workflow function. `.df` is data frame that combines the occurrence data and covariate data. `.df` is passed automatically in workflow from the process module(s) to the model module(s) and should not be passed by the user.
`features`	A length-one character vector (i.e. a string) defining the types of features to use for all covariates (see details).
`regmult`	A positive scalar giving the multiplier for the degree of regularisation. Higher values mean more regularisation.
`clamp_predictions`	Whether to clamp predictions when extrapolating.
`prediction_type`	The scale on which to make predictions (see details for options).

The maxnet R package fits MaxEnt models using the glmnet package, which enables efficient fitting of glms with regularization. Unlike MaxEnt, MaxNet does not require the user to download and install the MaxEnt java executable.

features should be a string including an 'l' for linear features, 'q' for quadratic features, 'h' for hinge features, 'p' for pairwise interactions and 't' for threshold features. E.g. to use only linear features and their interactions, either features = 'lp' or features = 'pl' would work. the default value of 'default' uses maxnet's default settings, adjusting the set up based on the number of occurrence records np: 'l' if np < 10; 'lq' if np < 15; 'lqh' if np < 80; or 'lqph' if np >= 80. I.e. the default never uses threshold features.

prediction_type corresponds to the type argument in the maxnet predict function. maxnet enables types: 'link', 'exponential', 'cloglog' and 'logistic'. However most output modules expect predictions to be made on the probability scale, for which only 'cloglog' and 'logistic' are guaranteed to work.

1.0

2016-12-21

presence/background

Nick Golding, nick.golding.research@gmail.com

maxnet::maxnet

Other model: BiomodModel, Domain, GBM, LogisticRegression, MachineLearn, MaxEnt, MaxLike, MyMaxLike, NullModel, OptGRaF, QuickGRaF, RandomForest, StochasticLogisticRegression, mgcv