Bias and Variance
Contents
Bias and Variance¶
The previous section introduced the basic concepts and techniques for building statistical models. This section focuses on how to assess whether a given model is a good model.
In creating a model, one specifies a plausible model to explain a data generating process and determines the most likely model, among all models of that kind, that explains the observed data. Fitting such a model raises a number of questions:
Does the fit model explain the observed data well?
Does the fit model explain the data generating process well?
Can any model explain the observed data well?
Is there another model specification that describes the data generating process more effectively?
These questions are answered using the concepts of bias and variance.
Content Summary¶
Evaluating the fit of a model.
Cross-validation.
Parameter searches.