2
$\begingroup$

I am analyzing a small dataset of 76 observations and I want to explore how 9 environmental predictors explain my response variable. For this I have decided to use regression trees because I am interested to see how the variables interact to each other

I am wondering if its right not to split the data to train & test dataset and just use the entire set to fit the model ?

IMPORTANT : My intention is NOT do use the model for prediction but just for exploration of my current data

I will appreciate any ideas/advice,

A

Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

Part of your exploration should be to test how well a tree can model your data and to do that you should use a train/test split. After you do that you can train your model with the entire dataset (or just training set) for further exploration.

Improve this answer
$\endgroup$
2
  • 1
    $\begingroup$ Thanks for your answer! So the split of the data is mandatory! I didn't quite understand what you're saying , So I split the data to train and test and then I train the model based on either train set or entire dataset? And after that I test the model based on the test data ? I am right? If that the case I can split 50/50 and train on the entire dataset and then test based on the 50% test subset right? Also what do you think about bagging afterwards as alternative to pruning the single tree? $\endgroup$
    – Anna Antonatou -Pappaioannou
    Commented Aug 2, 2021 at 11:50
  • $\begingroup$ Yes you are right. Since you have a small amount of data, if I were you, I would run leave-out-out cross validation to test how good your model is. You should do this because if your error is extremely high, any interpretation you do on your model would be basically meaningless. After you do that, I would train your model on your entire dataset and interpret the model. $\endgroup$
    – Joe Janssen
    Commented Aug 3, 2021 at 19:23

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.