I am writing thesis about predicting which corporate takeover targets attract a competing bidder. I first estimated a model using a logistic regression. The sample consists of 1350 single-bidder targets and 51 multiple-bidder targets.

I want to test the predictive ability with a holdout sample and use this prediction to construct a portfolio that generates more returns than the market (CAR); however this holdout sample has 250 single-bidder targets and only 13 multiple-bidder targets.

  1. Does it matter for the holdout sample that the number of multiple-bidder target is so small?
  2. Would a "leave-one-out-cross-validation" (LOOCV) be superior to the holdout sample to test the predictive ability? i.e. not model accuracy
  3. How I can do a "leave one out cross validation" (LOOCV) using spss?
  • $\begingroup$ I don't understand what the "i.e. not model accuracy" is doing in 2. It doesn't seem related to anything. Can you clarify or delete that? $\endgroup$ – gung - Reinstate Monica Jun 13 '14 at 15:22
  • $\begingroup$ thnax for your answer my question is about the way that i can use Leave-one-out cross validation to validate a simple linear regression $\endgroup$ – user48331 Jun 13 '14 at 17:15
  • $\begingroup$ Yoy seem to need an advanced analysis. You should definitely switch to some better language than SPSS, such as R. Also, leave-one-out is probably not good to test prediction ability, it perturbs the model to little. $\endgroup$ – kjetil b halvorsen Jun 14 '14 at 12:38

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.