LEGIT (version 1.4.1)

best_model.elastic_net_var_select: Best model from elastic net variable selection

Description

Best model from elastic net variable selection (based on selected criteria)

Usage

# S3 method for elastic_net_var_select
best_model(object, criterion, ...)

Value

Returns the best IMLEGIT model resulting from the glmnet path with associated information.

Arguments

object

An object of class "elastic_net_var_select", usually, a result of a call to elastic_net_var_select.

criterion

Criteria used to determine which model is the best. If search_criterion="AIC", uses the AIC, if search_criterion="AICc", uses the AICc, if search_criterion="BIC", uses the BIC, if search_criterion="cv_R2", uses the cross-validation R-squared, if
search_criterion="cv_AUC", uses the cross-validated AUC, if search_criterion="cv_Huber", uses the Huber cross-validation error, if search_criterion="cv_L1", uses the L1-norm cross-validation error (Default = "AIC"). The Huber and L1-norm cross-validation errors are alternatives to the usual cross-validation L2-norm error (which the \(R^2\) is based on) that are more resistant to outliers. For all criterion, lower is better, with the exception of search_criterion="cv_R2" and search_criterion="cv_AUC".

...

Further arguments passed to or from other methods.

References

Alexia Jolicoeur-Martineau, Ashley Wazana, Eszter Szekely, Meir Steiner, Alison S. Fleming, James L. Kennedy, Michael J. Meaney, Celia M.T. Greenwood and the MAVAN team. Alternating optimization for GxE modelling with weighted genetic and environmental scores: examples from the MAVAN study (2017). arXiv:1703.08111.

Examples

if (FALSE) {
N = 1000
train = example_3way(N, sigma=1, logit=FALSE, seed=7)
g1_bad = rbinom(N,1,.30)
g2_bad = rbinom(N,1,.30)
g3_bad = rbinom(N,1,.30)
g4_bad = rbinom(N,1,.30)
g5_bad = rbinom(N,1,.30)
train$G = cbind(train$G, g1_bad, g2_bad, g3_bad, g4_bad, g5_bad)
lv = list(G=train$G, E=train$E)
fit = elastic_net_var_select(train$data, lv, y ~ G*E)
summary(fit)
best_model(fit, criterion="BIC")
 # Instead of taking the best, if you want the model with "Model index"=17 from summary, do
plot(fit)
# With Cross-validation
fit = elastic_net_var_select(train$data, lv, y ~ G*E, cross_validation=TRUE, cv_iter=1, cv_folds=5)
best_model(fit, criterion="cv_R2")
# Elastic net only applied on G
fit = elastic_net_var_select(train$data, lv, y ~ G*E, c(1))
# Elastic net only applied on E
fit = elastic_net_var_select(train$data, lv, y ~ G*E, c(2))
# Most E variables not removed, use lambda_mult > 1 to remove more
fit = elastic_net_var_select(train$data, lv, y ~ G*E, c(2), lambda_mult=5)
# Lasso (only L1 regularization)
fit = elastic_net_var_select(train$data, lv, y ~ G*E, alpha=1)
# Want more lambdas (useful if # of variables is large)
fit = elastic_net_var_select(train$data, lv, y ~ G*E, n_lambda = 200)
}