From the CFAI sample exam: An assumption of classical linear regression is: a) independent variable is random b) expected value of the error term is 1. c) error term is correlated across observations d) variance of the error term is the same for all observations.

I would go with D.

my thoughts: a) why cannot the independent variable follow a random walk - y1=y0+error term? b) usually epsilon is distributed with e~(0,sigma^2) c) autocorrelation - better not… d) yes the variance has to be the same… I’m struggling with a) … why cannot a) be the answer?

D for me too…homoskedastic…

Answer should be D A) independent variable is NOT random B) E(Epsilon) = 0 C) Corre(Epsilon)(obs1, obs2) = 0 D) Variance(Epsilon) = CONSTANT -Dinesh S

D is correct. this is very typical CFAI quant- just test you on a definition, no calculator needed. Anyone take the BSAS practice test yet- define heteroscedasticity? Alex, I will go with what is the opposite of answer D…

“We assume the independent variable is not random, but that assumption is clearly often not true. It is unrealistic to assume the monthly return of the S&P 500 are not random. Does that mean the regression model is incorrect? Fortunately, No, Econometricians have shown that even if the independent variable is random, we can still rely on the results of regression models given the crucial assumption that the error term is uncorrelated with the independent variable.” CFA Notes Book 1 So, basically its an assumption that we are supposed to follow in classical linear regression.

If the independent variable is random, the correlation r will be zero. the Se should be random, independent from the linear relationship with X or Y.

thanks for the explanation guys

CFALondon0109 Wrote: ------------------------------------------------------- > Fortunately, No, Econometricians have shown that > even if the independent variable is random, we can > still rely on the results of regression models > given the crucial assumption that the error term > is uncorrelated with the independent variable." > Econometricians?!

They’re like mathemahoojibars.