I actually have a a dataset of 1100 observations containing 99 features and one targeted value (Y). The idea is that my dataset is split in two: 100 observations where the targeted value is know and 1000 observations where the targeted value is unknown.

I have chosen my model based on the 100 sample observations but know I am asked to: "Estimate the predictive performance of your model. We are interested in the root mean squared error. As you do not know the true values y, you cannot just calculate the error, you need to estimate it".

I have no idea of how to compute the MSE since I only have $\hat{y}$ and $y$ is missing...

If someone has any clue to give me, I would be grateful..

  • $\begingroup$ I'm confused. In your opening sentence, you indicated you had "99 features and one target value ($y$)". But then in your penultimate sentence, you indicated you don't have $y$. Which is it? $\endgroup$ – StatsStudent Mar 21 at 14:17
  • $\begingroup$ In my first 100 observations $y$ is known and missing for the last 1000 ones. $\endgroup$ – Jules Mar 21 at 14:39
  • $\begingroup$ I see. You can still compute the MSE on the 100 known values. $\endgroup$ – StatsStudent Mar 21 at 14:55

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.