constructGradient {Hmsc}R Documentation

constructGradient

Description

Constructs an environmental gradient over one of the variables included in XData

Usage

constructGradient(
  hM,
  focalVariable,
  non.focalVariables = list(),
  ngrid = 20,
  coordinates = list()
)

Arguments

hM

a fitted Hmsc model object

focalVariable

focal variable over which the gradient is constructed

non.focalVariables

list giving assumptions on how non-focal variables co-vary with the focal variable or a single number given the default type for all non-focal variables

ngrid

number of points along the gradient (for continuous focal variables)

coordinates

A named list of coordinates were model is evaluated in spatial or temporal models. The name should be one of the random levels, and value can be "c" for mean of coordinates (default), "i" for infinite coordinates without effect of spatial dependence, or a numeric vector of coordinates where the model is evaluated.

Details

In basic form, non.focalVariables is a list, where each element is on the form variable=list(type,value), where variable is one of the non-focal variables, and the value is needed only if type = 3. Alternatives type = 1 sets the values of the non-focal variable to the most likely value (defined as expected value for covariates, mode for factors), type = 2 sets the values of the non-focal variable to most likely value, given the value of focal variable, based on a linear relationship, and type = 3 fixes to the value given. As a shortcut, a single number 1 or 2 can be given as a type used for all non-focal variables. If a non.focalVariable is not listed, type=2 is used as default. Note that if the focal variable is continuous, selecting type 2 for a non-focal categorical variable can cause abrupt changes in response.

The function needs access to the original XData data frame, and cannot be used if you defined your model with X model matrix. In that case you must construct your gradient manually.

Value

a named list with slots XDataNew, studyDesignNew and rLNew

See Also

plotGradient, predict.

Examples

# Construct gradient for environmental covariate called 'x1'.
Gradient = constructGradient(TD$m, focalVariable="x1")

# Construct gradient for environmental covariate called 'x1'
# while setting the other covariate to its most likely values
Gradient = constructGradient(TD$m, focalVariable="x1",non.focalVariables=list(x2=list(1)))


[Package Hmsc version 3.0-13 Index]