Discussion questions#
We discuss in lecture that when researchers build a model, they often say they are doing either prediction (forecasting outcomes) or inference (understanding relationships). Can you think of situations where a model that predicts well also helps us understand why something happens and situations where strong prediction actually prevents interpretation?
One person in your study performs far better than everyone else. If you include them, your study shows an effect. If you remove them, the effect disappears.
What might explain this person’s score?
When is it fair to remove someone from a dataset?
Is this person a problem, or is your explanation of the data the problem?