PIMS-SFU Computational Math Seminar: Amanda K. Criner

Abstract: Zero mean, identically, independently distributed (i.i.d.) error is implicitly assumed in many data analysis algorithms. These assumptions are explained in example simulated data sets. Strategies to recognize violations in these assumptions are demonstrated on the example data sets. The consequences of violation of each of these assumptions, bias, overconfidence and underconfidence, are discussed. Finally, methods and resources for accommodating the violation of the zero mean i.i.d. error assumptions and extensions are demonstrated on the example data sets.