Model selection

Bayesian statistics

How do we select the best model? What do we mean by the “best”? We cannot simply use the accuracy of the model because of Bias–variance tradeoff. Akaike information criterion and similar measures penalizes the number of parameters.

In more traditional Regression analysis, several independent variables are sequentially introduced to build increasingly complex models and Akaike information criterion (or similar) is used for the model selection.


General references

Network science