Open nickeubank opened 9 months ago
I Agree. You can correctly predict using correlations. That said, the absence of causality makes it likely that your prediction goes awry -- think about the dumb predictions about Washington Football team and presidential elections. You do not need causality to make a prediction, but causality helps ensure that the prediction is not simply spurious. We evaluate predictions by the ability to classify correctly (cross-validation, etc.) -- not explain why an outcome occurs.
Adriane: