Sometimes the best model is no model
Data can be used to discover relationships between predictands and an assumed set of predictors, even in the absence of any prior theory. However lack of an underlying theory imposes a cost: the model-building procedure…