ECON 104 Lecture Notes - Lecture 23: Multicollinearity
Document Summary
It"s almost equal to r2 except for you multiply the second term with an object. Removing k from equation, it will be r2. If k is not 0, the object will be smaller than 1. As k increases, adjusted r2 will go down. Tradeoff of adjusted r2, ratio ssr/tss will decrease. As you increase number of regressors in equation, n-1/n-k-1 will go up, and thus adjusted r2 will compare models and look at if diff ssrs compensate for different n-1/n-k-1. If large sample size/ n>>k will pretty much be 1, and so changes in k will make super small change. R2 is a measure of how well you can guess different observations. Adjusted r2 corrects for problem of having more than 1 regressor. If include pctel -> predictions for test scores will be more precise than it used to be (that"s what.
