Model Checking Flashcards
(10 cards)
What is the main difference between model selection and traditional hypothesis testing?
Model selection compares many models at once to find the best balance of fit and simplicity, while traditional hypothesis testing compares one model against a null hypothesis.
What does AIC measure?
AIC measures the quality of a model based on its goodness of fit and complexity (number of parameters).
Why is a lower AIC better?
Because it indicates a model that fits the data well with fewer parameters, avoiding overfitting.
What does ΔAIC tell you?
ΔAIC is the difference between a model’s AIC and the best model’s AIC, indicating how close or far it is from the best model.
What is model averaging, and when should you use it?
Model averaging combines predictions or estimates from multiple models weighted by their AIC weights, used when no single model is clearly the best.
What is the purpose of log-likelihood in model comparison?
Log-likelihood quantifies how well the model explains the data and is used in calculating AIC.
If two models have a ΔAIC less than 2, what should you consider?
Consider that both models have similar support and it might be better to use model averaging.
What does AIC penalize, and why?
AIC penalizes models with more parameters to avoid overfitting.
What is an advantage of using all-subsets model selection?
It tests all possible variable combinations, ensuring you find the best model among all options.
How do AIC weights help in model interpretation?
They estimate the probability that each model is the best among the candidates, helping to weigh models in inference.