9 The world of statistical modeling

 

This chapter covers

  • The purpose and application of common classes of statistical models
  • Evaluating the information available when fitting a model
  • Fitting a statistical model to a dataset and iterating on it to improve performance
  • Explaining the results of statistical models and generating predictive deliverables

Three data analysts walk into a bar. The first says, “I bet I can figure out the top five traits that predict whether a person orders beer, wine, or spirits. We can use that model to better plan the inventory.” The second retorts, “Well, give me info on the next 100 patrons, and I’ll use your model to forecast all their orders in advance.” The third smirks, “Why wait? Give me real-time data, and I’ll use your model to predict a drink right as they’re about to order it. Now, that will really impress the patrons!”

What’s the difference between the type of model that each analyst is proposing and the other two? Are all three approaches valuable? Can they all genuinely use the same statistical model to predict the same phenomenon, with a different approach and desired output? And are the analysts actually proposing successively better alternatives?

9.1 The many faces of statistical modeling

9.1.1 Classes of statistical models

9.1.2 Model output and diagnostics

9.1.3 Exercises

9.2 The modeling process

9.2.1 Exploratory analysis

9.2.2 Fitting a model

9.2.3 Exercises

9.3 The statistical model and its value

9.3.1 Explanatory models

9.3.2 Predictive models

9.3.3 Exercises

Summary