Page 374 #14: For F-test, text says that null is that “all three population slope coefficients equal 0–that the three variables jointly are statistically not related to the returns.” Schweser describes F-test null as a test of whether “at least one of the independent variables explains a significant portion of the variation.” Isn’t there a discrepancy in the definition: “jointly related” vs. “at least one”? Page 389 #20: R^2 = 0.36, and answer says that this implies that squrt(0.36) = 0.6 = correlation. However, we learn that this is only the case with 1 independent variable. In this question, there are four independent variables, so I’m thinking that the relationship should not hold? Page 464 #13 & #14: Answers to both say that after adding a lagged dependent variable to correct for serial correlation, “if the residuals from the AR(2) model are serially uncorrelated we should then test for seasonality and ARCH.” If there is no serial correlation, how can there be seasonality? I thought seasonality is a type of serial correlation.
why did you think seasonality was serial correlation? seasonality is how e.g. in terms of Sales - sales of say the 4th quarter is significantly higher (due to the holidays) than say in early summer… It relates to value of the “y” variable due to change in the X value (typically Quarters would be dummy variables 1,2,3 and 4th quarters avg sales would be the b0 term). serial correlation relates not to the value of Sales - but to the error term.
i had always thought of seasonality as autocorrelation of the residuals of every 4 or 12 periods, since you test for it by testing the significance of the residuals at those periods. you’re saying that while autocorrelation is more of a pattern in residuals (i.e. positive correlation from period to period), seasonality is more of a pattern in the dependent variable (i.e. every 12th month)?