Variance of OLS estimators in the presence of serial correlation

[@jealey12]

83 Vids in about 24 hours, you're saving my degree!

[@jasonleewkd]

Hello Ben, could you tell me where this video fits in the whole scheme of your undergraduate course of econometrics? The past few videos (76 to this one) on derivations was pretty math heavy and I got lost in them.

[@aishwaryapotdar1348]

Hey Ben, wouldn’t the AR (1) covariance be (p^j * sigma2)/(1-p^2) ?

[@nkosiphilepamelazele887]

hello Ben, l need to understand how the residual variance is likely to under estimate the true variance

[@iamskaars]

Thank you so much for your help, You gave me the hope of passing my econometrics class! (I still have to watch the rest of the videos) I just want to know sir what is SSX ? And the we have to know all the proofs...

[@indragesink]

Ayoub Morghati SSX, I think, equals Sxx (with a typo).

[@aboyinfinland9230]

@@indragesink it should have been SST actually (total sum squares). I checked with Wooldridge book.

[@aboyinfinland9230]

@@indragesink kzbin.info/www/bejne/e5nZgZmQjq92oac Refer to this as well.

[@achwww]

Thanks so much for your video, but may I know why is that term always positive and >0? Can't it be negative in some cases?

[@miguelmarin9076]

It is not ALWAYS positive, but that is usually the behavior (your past observations usually have a direct relationship, and not inverse, to the present observation). As a matter of fact, if you have negative autocorrelation, the estimated relationship (beta) will look less significant due to the fact that the estimated variance is bigger than the real one