I’m confused on 2 ‘challenge’ questions in book 1 of schweser pg. 213. question 13 - What is the remedy for serial correlation? a) use robust standard errors b) use generalized least squares personally i thought both were used to correct for heteroskedasticity. i was looking for hansen method to be the answer. question 14 - Can adjusted R squared ever be higher than R squared if there 50 observations? Whille I figured this was unlikely, it didn’t seem like I could say never. Adjusted R squared can theoretically be higher than R squared, right? Thanks for the feedback! nerdattax
adjusted r-squared, mathematically, can never be higher than r-squared. -cfastudent, cfa
I’ll bet the book likes both a) and b) for 13. Generalized least squares is pretty clear when you have some nice correlation structure among the residuals and robust standard errors kinda work too (I don’t like this approach, btw). cfastudent is right about adj. r^2. and I don’t know what the Hansen method is for dealing with serial correlation. There is a Hansen test to see if you have it, but did Hansen propose some remedy for it?
I did this question today as well and also picked least squares as my answer but according to Schwesser the answer is robust standard errors only.
CFAI material also picks robust standard errors and the Hansen method as it also corrects for conditional heteroskedasticity… pp 303.