Retrain best model on full dataset in R - r

Question

I have two models to select from and using some criteria I choose one of the two. (The below is just an example, I know it doesn't make much sense)
library(forecast)
set.seed(4)
sample_dat= sample(1:nrow(cars), 5)
train = cars[-sample_dat, ]
test = cars[sample_dat, ]
models = list(lm(dist ~ speed, train), glm(dist ~ speed, train, family = "poisson"))
test_res = sapply(models, function(x) accuracy(predict(x, test, type = "response"), test$dist)[2]) #Getting the RMSE for each model
best_model = models[which.min(test_res)]
How can I retrain the best model using the full dataset (train + test)? I checked the update and update.formula functions but these don't seem to be updating the data part.

Answer 1

update(best_model[[1]],data = rbind(train,test))
You do not want to change the formula since that is the best model but rather update the data

Answer 2

Base R using your own logic, first creating a list mirroring the models list:
set.seed(4)
sample_dat= sample(1:nrow(cars), 5)
train = cars[-sample_dat, ]
test = cars[sample_dat, ]
models = list(lm(dist ~ speed, train), glm(dist ~ speed, train, family = "poisson"))
model_application = list(as.expression("lm(dist ~ speed, cars)$call"),
as.expression("glm(dist ~ speed, cars, family = 'poisson'))$call"))
test_res = sapply(models,
function(x){
# Store a function to caclulate the RMSE: rmse => function
rmse <- function(actual_vec, pred_vec){sqrt(mean((pred_vec - actual_vec)**2))}
# Getting the RMSE for each model: numeric scalar => .GlobalEnv
rmse(test$dist, predict(x, data = test, type = "response"))
}
)
best_model = models[[which.min(test_res)]]
applied_model <- eval(eval(as.expression(parse(text = model_application[[which.min(test_res)]]))))