Dear analyst forum members this might sound too early for many of you but for me given the spoce of CFA curriculum I need one year to prepare in a rithm I can manage well. This way I enrolled in the CFA Level 2 exam on the first day available ( 27th July), and I started studying already and have a doubt.
My doubt is as follows imagine a time series with data from 1993 to 2005, ln salest= 0.05+ 0.55 ln salest-1 , being salest the value of sales/ quarter and I want to estimate the values of sales for all quarters of 2006 and in the end I want to calculate the RMSE, and I have the real values for all quarters of 2006. what makes me confusion is this point in estimating the 2006 values using the regression for exemple for Q2sales I will use the predicted value for Q1 sales given by the regression previously, or the actual value of Q1 sales I have?
Thanks in advance
I think you should have the 2005 real quarter sales, not the 2006. The idea is to predict 2006 using your autoregressive regression equation. I would look over again. However, in case you got the 2006 real quarter sales, then the idea is to predict 2007 quarter sales.
Hope this helps!
thanks for your reply, if I understood well, what you are saying is that in order to calculate RMSE I need to have all the real data for that year and the estimated value for Q2 2006 by the regression is based on the real Q1 never on the predicted Q1 by the regression previously
Sorry, I misunderstood your question.
You are told to calculate the RMSE at forecasting 2006 using the AR equation, so the question provided 2006 real values, Ok.
You need to forecast the four quarters of 2006 using the AR equation ( ln salest= 0.05+ 0.55 ln salest-1 ). Once you got the four forecasted values, compare each one with the 2006 real values and calculate their difference (those differences are the errors of forecast). Then, sum them all, divide by n-k-1 and square root that ratio. You got the RMSE right there.
Hope this helps!
Ok, now I have my doubt answered, thanks