ISLP Flashcards
What are the tradeoffs when deciding on a model to use for prediction?
Prediction accuraucy vs interpretiabilly
Good fit vs over fit our underfit
Number of variables used parsimony vs black box
What is a parametric model approach?
This is where you first assume the functional shape eg.linear
We must then use the training data to fit the model
What does a non parametric model approach consist of?
This is where a functional form is not presumed and a model is used to fit to the training data.This can yield more accurate results but takes a lot of data and can lead to overfitting.
What is overfitting?
This occurs when a model learns the training data too well and is unable to generalise to new data.This can be caused by using too complex of a model or too little data.
Explain the trade off between interpretability and prediction accuracy?
As the flexibility of a model increases it is usually able to fit multiple shapes to the data.However, as this occurs the interpretabillity reduces due to a complex model and vice versa.Therefore, the choice of model depends on the goal of the model.
When are more/less flexiblile models desired?
More are usually desired when you are trying to predict the output of something.Less flexibile is desired when you are trying to make some sort of inference eg relationship between 2 variables.
What is the difference between bias and variance?
Bias is the difference between the expected value and the actual value it is squared to avoid being negative.Variance is the difference in f(x) in response to different test data.
How to calculate the Test MSE?
The test MSE is the sum of the variance of f hat of x0 plus the bias of fhatxo squared and variance epilson.
