One example in the section titled "Split into training and testing sets" would be to use the createDataPartition() function from the caret package for creating

One example in the section titled "Split into training and testing sets" would be to use the createDataPartition() function from the caret package for creating training and testing sets. He says "In R, there are packages to make sampling simpler, but aren’t much more concise than using the built-in sample function" but using caret is more concise. Added: Later in the section on random forests he says "With R, there are many smaller packages containing individual algorithms, often with inconsistent ways to access them." Which is why you want to use the caret package as it makes accessing many machine learning packages consistent and easy.