I'm analyzing the Titanic dataset, and I've been trying to understand the predictive power of the Age feature relative to passenger survival. My intuition is that younger people had a higher chance to survive because they were more resistant to cold water, or that older people survived because they were taken care of by younger ones.

While I can see no difference between the distributions of the two populations, survivors and victims:


the permutation importance for my ‎XGBoost model suggests that the Age feature has a certain importance in predicting whether someone survived or not.

permutation importance

Is this difference coming from the fact that:

  1. The predictive power is so small it cannot be observed from a violin plot?
  2. Age alone is not enough to predict survival, but only in combination with other features?
  3. I am misinterpreting the violin plot?

I know it might be impossible to say without having the dataset on hand, so I'm curious what approaches I could try to use to find the reason behind this difference.


Summary plot suggests that:

  • Higher age is associated with not surviving
  • Population of younger people (lower age) is smaller than that of older people (there are a few blue spots on the right)



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.