CFA Level 2 Exam-tips Blog Let's Do the Time [Series] Warp Again! By Jodi Joachim, CFA - Director of Content Development

Hello Level 2 candidates!
I cannot believe that it has been almost three weeks since I last blogged! I am usually too prolific here...it's just been a bit nuts around here, with getting questions ready for the Weekly Online Program and answering participant's questions and working on Secret Sauce (there are so many changes at Level 2 this year).  I have been ignoring the blog, so my apologies!

Those of you who are fans of The Rocky Horror Picture show can tell my subject line is a play on one of the songs in the play. It was too good to pass up, because, as many of you have probably already figured out, you will have to read the topic review on Time Series in Quantitative methods more than once to "get" it. The material was in Level 3 when I studied, I was on vacation at the time and I remember sitting at the UNLV library staring out the window while my friends took in Las Vegas while I was doing what I could to study and going nuts trying to sort it out.

But, I did, and I think it took 3 or 4 reads. So, my main point is that don't panic if it doesn't sink into your brain the first time. Work problems. Make your own notes on this one. And here are some points to keep in mind as you review the material:

• For a times series, step one is to look at the data and try to figure out what time series to use. Then, you have to look at the residuals and look for a trend or pattern. The "residuals" are where a lot of the "action" is. Always remember to check the residuals to see if they indicate any problems. (For example, statistically significant lags).
• If you have a model that is not covariance stationary, you can still use it if you correct for the problems, for example, by first differencing. So, while the time series may not be "OK" at the first pass, you can "correct" it and then use it later.
• So many candidates get confused when they see an AR model with what seems like 2 "lags" but it's really an AR(1) with a seasonal lag.  What that means is that we "corrected" an AR(1) model with a seasonal lag component. We still have to check for covariance stationarity, etc, and may have to add additional lags to correct the times series.

Good luck with the material,
Jodi