ECON 104 Lecture Notes - Lecture 12: Regression Analysis, Loss Function, Prior Probability
Document Summary
Slide 13 + 14: bayes estimators, bayesian stats is like a switch - opposite of bayesian stats is frequentist stats, bayesian stats ask more philosophical questions about what is randomness c) P(head) = but outcome is random/ not probabilistic/ not realised. Compute expected/ integrated e(l(theta, a (x))) and find ___ that ___ Ridge estimator used in lots of machine learning. Estimator for minimising squared loss, minimise squared distance between parameter. To minimise population mean by minimising square distance between mean and data points, theta that minimises this is the sample average. These estimators differ because when you have multiple thetas to estimate, having too many parameters to estimate given sample size, these estimators force us to pick some thetas that are small. Don"t estimate those ones precisely but focus on remaining parameters instead. As theta ->, lambda*theta^2 -> 0 to determine educational outcomes but don"t know ___ these estimators will force.