chapter 8 Flashcards
statistical inference
a collection of methods that deal with drawing conclusions from data that are prone to random variation
central idea: probabilistic/statistical model: view data as realisations of random variables
statistical model
gives the probability distribution under which we assume the data have been generated. Typically this model contains unknown parameters
Frequentist approach
parameters are considered fixed quantities
Bayesian approach
parameters are assigned a probability distribution that quantifies information we have about them
undefrit models
experience high bias—they give inaccurate results for both the training data and test set
overfit models
experience high variance—they give accurate results for the training set but not for the test set
bias
average distance between predictions and the truth
bias-variance trade off
high bias ⇒ underfiting on training data
high variance⇒overfitting on training data
low bias ⇒ adapt well on training data
low variance ⇒ generalise well on unseen data
high bias + high variance
high average training error, inconsistency in prediction
high bias + low variance
high average training error, consistent
low bias + high variance
low average training error, inconsistency in prediction
low bias + low variance
low average training error, consistent
decision rule
reject H0 for large values of T; so if T ≥ c. T € K := [c, ♾️)
acceptance region
the set of values t, for which H0 is accepted, so K^c
rejection region
the set of vales t, for which H0 is rejected, so K
