0

I am studying for the Statistical learning exam and my professor has put me in doubt as to when it is preferable to use cross-validation and when the validation set: he said that with less than 2000 observations it is better to use cross-validation, with more than 2000 observations it is better to use the validation set. Yet I think it is more the other way round, also because of the more stable estimates generated by the cross-validation method.
Can anyone help me?
Thanks

Enrico Caria
  • 1
  • 2
    I don't like this rule - but anyway the logic goes along these lines: With large data, you can usually spare 20% and still have enough training rows. The smaller the data, the more it hurts to reduce it by 20%. – Michael M Feb 13 '23 at 18:48
  • 1
    I use cross validation on around 100 observations samples too. – Aksakal Feb 13 '23 at 18:49

0 Answers0