Properties of leverage points in regression with proofs - note typo
HTML-код
- Опубликовано: 16 сен 2024
- The properties of leverage points in regression are:
sum of leverages over observations equals the number of parameters in the model (intercept plus number of slope parameters)
TYPO: Var(residual) = sigma^2(1-h_ii) not sigma^2sqrt(1-h_ii)
each leverage point takes a value between 0 to 1 inclusive.
I present the proofs. Along the way I show you that positive semi definite matrix has non negative diagonal elements.
thanks! This is so much better than the tutor Phil who teaches me ST300!
absolute banger
Shouldn't the variance of residuals be equal to sigma ^2 * (1-hi) ?
Yes, thanks. At 1:17 the variance of the residual of the ith point is sigma^2*(1-h_ii). Basically, I should not have a square root on the screen over 1-h_ii
ALSO
at 11:12 on the 4th line, the equal sign should be pointing at e'e as it's e'e=1
this is great, thank you!
Hi your I liked your video. I had a question though. My homework states if n=p show that H is the PxP identity matrix. What is his and Yi_hat. Our teacher never taught us this and I was wondering if you could show this?
If n=p then X is square and if it's invertible then what you say is true - that the hat matrix is the identity matrix. This is just a curiosity, not a result of practical interest as mostly we have n > p in courses in basic regression, (or p