I am trying to build a model with binary output and when used Logistic regression, I found none of the 20 variables(with about 5 factor variables) as significant (p-value higher than alpha=0.05). But when I used the same dataset for XGBoost Model it gave 90% accuracy on test set in R.

Is this normal?. Am I doing right?. I have only 40 observations, and split the training and test sets in the ratio of 75:25%. Does this is any way issue for this difference in the results between 2 models.

What can be done further to have a robust model?. I can think of getting more sample data only.

Any other ideas?

  • 2
    $\begingroup$ I'm voting to close this as off-topic because it's unclear why you would expect two completely different models to yield similar results. $\endgroup$ – Sycorax Jul 3 '18 at 1:34
  • $\begingroup$ On top of Sycorax's comment, there is no reason why descriptive statements like accuracy should have anything in relation with inferential statements like a p value. $\endgroup$ – Michael M Jul 3 '18 at 7:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.