Statistical Error and Predictive Model Validation

How accurate is a model for predicting your shoe size based on your height? What tools can you use to assess how far off these predictions are? What are the pros and cons of having less error in a predictive model? It is very rare to create a perfect predictive model, so how do you know whether to trust the models you create? In M214 Statistical Error and Predictive Model Validation, you will explain and quantify variation and its sources while mastering multiple methods of model validation, uncertainty, and comparison to inform decisions. You will utilize technology to practice statistical techniques to improve your predictive models such as F-tests, sensitivity analysis, transforming variables, adding or excluding variables, ANOVA, residual plots, normalizing variables, or cross validation. You will learn about the balance between bias and variance, and that a model with low variance does not necessarily equate to the best possible model. Your ultimate goal will be to answer the question “What might happen in the future?” based on prior knowledge—e.g., your data. You may use this information to diagnose a problem and draft its solution. Statistical error and predictive model validation are useful for careers in a variety of fields, like actuarial science, engineering, public health, and environmental sciences.