Be careful regarding subtle points in Quants following, regarding, through, careful, standard

zbird2134

I was just going through my Level II stuff and found an interesting concept, which was missed out in Schweser notes. Make note of the following:

1. In case of Multiple Regression, regression coefficient (b1) is not affected by serial correlation but standard error of regression coefficient (sb1) becomes inconsistent due to correlation of residuals (serial correlation).

2. In case of serial correlation in Autoregressive (AR) Model, regression coefficient (b1) as well as standard error of regression coefficient (sb1) become inconsistent.

Any comments will be highly appreciated.

mp3bu

I can't believe that Level II requires you to even know what "inconsistent" means in this context. Anyway, 1) Is only sort of true. Suppose that you are estimating the equation y = a + b x and you estimate a and b using OLS. Your estimators are surely consistent for a and b in the sense that if you gather enough data your estimators will get closer and closer to a and b. The problem is that the serial correlation tells you that your model is wrong because you are leaving out whatever is causing the serial correlation. I think telling someone your estimators are consistent, but your model is wrong is pretty foolish.

If you have serial correlation in the residuals, you need to figure out what's up and fix it, consistency or no.