Independence of Errors

Assumptions: Errors of prediction for each case (participant) have to be unrelated to the errors of prediction for all the other cases (participants). This means no auto-correlation, groupings, or other things that cause cases to be related to one another in ways not captured by the model.