Prediction models

Slides

• “Explain vs predict”
• “Prediction and overfitting”
• “Confidence and prediction intervals”
• “Cross validation”

Control questions:

• What is your goal if you are concerned about model assumptions being fulfilled?
• What is your goal if you are mostly concerned with out-of-sample performance?
• What is overfitting and how can we avoid it?
• Which is widest - confidence interval of prediction interval?
• Why do we need a train-test split?
• What is the end-goal of a cross validation routine?