Retrieve models from selection table

Generate or extract a list of fitted model objects from a "model.selection" table or component models from the averaged model ("averaging" object), optionally using parallel computation in a cluster.

get.models(object, subset, cluster = NA, ...)

Arguments

object: object returned by dredge, model.sel or model.avg.
subset: subset of models, an expression evaluated within the model selection table (see ‘Details’).
cluster: optionally, a "cluster" object. If it is a valid cluster, models are evaluated using parallel computation.
...: additional arguments to update the models. For example, one may want to fit models with REML (e.g. argument REML = TRUE in some modelling functions) while using ML for model selection.

Value

list of fitted model objects.

Details

The argument subset must be explicitely provided. This is to assure that a potentially long list of models is not fitted unintentionally. To evaluate all models, set subset to NA or TRUE.

If subset is a character vector, it is interpreted as names of rows to be selected.

Note

"model.selection" tables created by model.sel or averaged models created by model.avg from a list of model objects (as opposed to those created with model selection tables) store the component models as part of the object - in these cases get.models simply extracts the items from these lists. Otherwise the models have to be fitted. Therefore, using get.models following dredge is not efficient as the requested models are fitted twice. If the number of generated models is reasonable, consider using lapply(dredge(..., evaluate = FALSE), eval), which generates a list of all model calls and evaluates them into a list of model objects.

Alternatively, getCall and eval can be used to compute a model out of the "model.selection" table (e.g. eval(getCall(<model.selection>, i)), where i is the model index or name).

pget.models is still available, but is deprecated.

Author

Kamil Bartoń