Skip to contents

Selects a best subset of predictor variables.

Source: R/select_pred.R
select_pred.Rd

Selects among a set of covariates the best set of npred predictors for a given response trait resp based on AIC values.

Usage

select_pred(.data, resp, covariates = NULL, npred)

Arguments

.data

A data frame with the response variable and covariates.

resp

The response variable.

covariates

The covariates. Defaults to NULL. In this case, all numeric traits in .data, except that in resp are selected. To select specific covariates from .data, use a list of unquoted comma-separated variable names (e.g. traits = c(var1, var2, var3)), an specific range of variables, (e.g. traits = c(var1:var3)), or even a select helper like starts_with("N").

npred

An integer specifying the size of the subset of predictors to be selected

Value

A list with the following elements:

  • sel_mod An object of class lm that is the selected model.

  • predictors The name of the selected predictors.

  • AIC The Akaike's Information Criterion for the selected model.

  • pred_models The Akaike's Information Criterion and the predictors selected in each step.

  • predicted The predicted values considering the model in sel_mod.

Author

Tiago Olivoto tiagoolivoto@gmail.com

Examples

# \donttest{
library(metan)
mod <- select_pred(data_ge2, resp = PH, npred = 10)
mod$predictors
#>  [1] "EH"   "EP"   "CDED" "KW"   "NR"   "CL"   "ED"   "NKE"  "NKR"  "EL"  
mod$AIC
#> [1] -628.8605
# }