k) Explain the types of heteroskedacity and how heteroskedacity and serial correlation affect statistical inference.
l) Describe multicollinearity, and explain its causes and effects in regression analysis.
It doesn’t really explicitly state that we have to know how to test and correct for them. Taking heteroskedacity as an example, there is a section in the book for ‘4.1.1 The Consequences of Heteroskedacity’. This should cover the LOS as described above. There are then separate sections for testing and correcting, 4.1.2 and 4.1.3 respectively.
I had skipped the testing/correcting but now I’ve come to the EOCs and Question 9B P.380 is asking me to describe in detail how I would formally test for conditional heteroskedacity.
Thoughts? Did I jump the gun? Do I have to go back and learn all that stuff?
I don’t think there’s a reason not to know it, from my personal opinion. It’s in the CFAI books and not an optional segment, so I think you should know it from their perspective as well.
There’s no point to learn statistics (or any topic, really) if you’re not going to learn what you’re actually doing and if you’re doing it properly (fitting and building a model, testing assumptions, correcting for violated assumptions)…Unless you just want to sound impressive at the holiday party when you use vocabulary words like autoregressive conditional heteroskedasticity or multicollinearity, just hope that no one else knows what they’re talking about either.
I would highly recommend that you a) learn how to detect the problem; b) know what the impacts to the model are if one of the above-mentioned are present; c) how to correct the problem(s)
It’s all fair game but (like I just said in another post) quant is not heavily weighted so don’t get too caught up in it. I thought quant was the hardest read in all of L2. Could’ve spent 6 months studying just that topic.