Stepwise AIC using forward selection in R

Question

I am trying to do a forward variable selection using stepwise AIC in R but I don't think that I am getting the desired results. Specifically, the function should start with no variables and keep adding variables and get their AIC values. However, when I run this I only get an AIC value for all variables. Where am I going wrong? here is my code-


model.full <- lm(distance ~ ., data = FAA_unique_without_speed_air)
model.null<-lm(distance ~ 1,  data = FAA_unique_without_speed_air)
modAIC <- MASS::stepAIC(model.full,direction='forward', scope=model.full, k = 2)

output -

Start:  AIC=9161.49
distance ~ aircraft + duration + no_pasg + speed_ground + height + 
    pitch

could we please have a [mcve] ? – Ben Bolker May 20 '21 at 16:03 — Ben Bolker, May 20 '21 at 16:03

score 3 · Accepted Answer · answered May 20 '21 at 16:20

I think it would be best to be explicit with the arguments of stepAIC, rather than use the defaults. Try:

Provide the null model as the initial model object when you want to do forward selection.
Provide both a lower and upper search formula in the scope.

For example, using the iris dataframe from the base library datasets:

library(MASS)

model.full <- lm(Sepal.Width ~ ., data = iris)
model.null <- lm(Sepal.Width ~ 1, data = iris)

MASS::stepAIC(model.null, direction = "forward", scope = list(lower = model.null,
                                                              upper = model.full))

Or if you want to use the defaults then you should be explicit about the default upper components included in the model:

stepAIC(model.null, direction = "forward", scope = ~ Sepal.Length + Species + Petal.Length)

However, as mentioned by @BenBolker you should post a reproducible example with your data so we can confirm.

Stepwise AIC using forward selection in R

1 Answers1