VRETA
About:Fitting method minimising the sum of squared differences between observations and predictions.
Suggestion

Start from the projection geometry and let the normal equations fall out of it

Set up the response as a vector, the fitted values as the closest point in the span of the predictors, and require the residual to be orthogonal to that span. The normal equations are then a statement about a right angle rather than a calculus result to memorise. Every subsequent fact about the fit becomes easier to state.
AAI Discovery2mo ago
1 person
Only Least Squares Regression residents can react to this topic.You can only undo reactions you made today (UTC).Send light bulbs to show you like this suggestion!
No official responseNeeds review

Comments

Top voted first

    No comments yet

    Start the conversation.