FlorenceCC
Member
Hi,
I have a few questions following my reading of the syllabus related to Diebold Chapt 5 and 6, as detailed below:
(1) linear vs. non linear trends. I seem to remember when reading the Stock and Watson syllabus that the concept of linearity applied to the parameters (i.e. B0, B1, etc.). Reading Diebold, it seems here that the concept of linearity/non linearity applies to the Time (t) variable. What drives the distinction between both, since we still seem to be applying some of the methodology and concepts introduced earlier (e.g. OLS method, SE of the regression, etc.)? And in addition, is it possible to apply the OLS method to non linear trends?
(2) when we talk about model selection criteria (s^2, AIC, SIC), we refer to k as the number of degrees of freedom: what do they represent in this context: the number of parameters included in our model (i.e. B1, B2, B3, etc.)? And is the intercept included in "k"? Again, the S&W syllabus introduced the distinction between the parameters for the independent variables and the intercept.
(3) With regards to seasonality, what is the driver in either choosing (i) s = n dummy variables or s=n-1 dummy variables and an intercept? In other words, why/when does it make more sense to omit a season and turn it into a reference point?
(4) I am not sure I understand how the trading day variations dummy work if it is supposed to represent the number of days? i.e. when does it take the values 0 and 1? -> I will edit this question to indicate to anyone who is interested, that it is actually answered when doing the "P1.T2.701. Regression analysis to model seasonality" question 701.3.
Many thanks in advance for your feedback!!
Florence
I have a few questions following my reading of the syllabus related to Diebold Chapt 5 and 6, as detailed below:
(1) linear vs. non linear trends. I seem to remember when reading the Stock and Watson syllabus that the concept of linearity applied to the parameters (i.e. B0, B1, etc.). Reading Diebold, it seems here that the concept of linearity/non linearity applies to the Time (t) variable. What drives the distinction between both, since we still seem to be applying some of the methodology and concepts introduced earlier (e.g. OLS method, SE of the regression, etc.)? And in addition, is it possible to apply the OLS method to non linear trends?
(2) when we talk about model selection criteria (s^2, AIC, SIC), we refer to k as the number of degrees of freedom: what do they represent in this context: the number of parameters included in our model (i.e. B1, B2, B3, etc.)? And is the intercept included in "k"? Again, the S&W syllabus introduced the distinction between the parameters for the independent variables and the intercept.
(3) With regards to seasonality, what is the driver in either choosing (i) s = n dummy variables or s=n-1 dummy variables and an intercept? In other words, why/when does it make more sense to omit a season and turn it into a reference point?
(4) I am not sure I understand how the trading day variations dummy work if it is supposed to represent the number of days? i.e. when does it take the values 0 and 1? -> I will edit this question to indicate to anyone who is interested, that it is actually answered when doing the "P1.T2.701. Regression analysis to model seasonality" question 701.3.
Many thanks in advance for your feedback!!
Florence
Last edited:
