Week 10 - Distributions, inference, diagnostics, Decision trees.

Question 1

What is R^2?

Answer 1

R squared is the proportion of deviance explained by the model.

Question 2

What is the difference between deviance and null deviance?

Answer 2

Null deviance explains how well y is predicted by model that uses an intercept only. The Deviance (or residual deviance) is a measure of goodness of fit. We want deviance to be low and Null deviance to be high.

Question 3

What are AIC and BIC?

Answer 3

They measure the model. Lower is better and BIC penalises model with higher variables more than AIC.

Question 4

What are regression trees?

Answer 4

They recursively partition data. Use average response value of each partition and estimate model. I.e. choose best partition by optimising criteria (e.g. sum of sqaure resids). It will choose where to split.