The pass mark from the Fall 2015 exam was released today. It was 52.5/88, which is right at 60%. That’s about what I was expecting.
This makes it a good day to spend some time going over that exam to see how we should manage our preparation time for the Spring 2016 sitting. I’ve already covered section D, on GLM’s, in the video marked D.0.1 How to Handle GLM Spring 2016. I’ll talk about the rest of the seminar just down below.
Before I begin, I want to point out that the lessons in this seminar were designed to offer complete and detailed coverage of the syllabus topics and the source readings. As always, some of those topics will be more popular than others, and some parts of the source readings will never be tested at all. Until several editions of the exam have passed, it will be difficult to suggest with any great confidence which lessons in the seminar are most important.
Thoroughly learning everything that could be on the exam will give you the best possible chance of passing the exam – but it will also take a very long time. If you try to fully prepare for everything that could potentially appear but you run out of time, you might end up failing to master some of the topics that are most likely to appear – and that could lead to disaster.
What I’m trying to say is, take the following recommendations as more of a list of topics that you should be absolutely certain to master than a list of things that you can skip.
Use the following guidance regarding which sections to spend the most time on in the full knowledge that I do not write any of the exam questions for this exam, I do not know what will actually appear on the exam, and I am not guaranteeing that the following guesses will be perfectly accurate.
If you want to follow along with the problems on the Fall 2015 exam, there is a lesson map posted in the Sample exams alongside the Fall 2015 exam.
B.1 Poisson Processes (Syllabus A.1-A.3, 0 to 15%)
Fall 2015: 4 or 5 problems depending on how you count, 9-11% of the total exam
Problem 1, 3, and 4 are all very predictable, much like old problems.
Problems 2 and 5 tell us that the source materials have changed, as problems like these didn’t appear on recent exams from 3L and ST.
Problem 2 could have been treated in B.1, but I chose to put the necessary topic for it in C.3.2 where we do order statistics.
Problem 5 is an application of the Memoryless property, which is in B.1.1.
It seems as though it is likely that we will regularly see exam questions from all of the lessons in section B.1, and from C.3.2 slide 4. I expect 2 or 3 of the problems to be very similar to the older problems, and 1 or 2 to be drawn more from the new source readings.
B.2 Continuous/Multiple lives and decrements (Syllabus A.4-A.5, 4 to 16%)
Fall 2015: 2 problems, 4.5% of the exam.
Both of the problems from this section avoided much of the material from the source readings and the syllabus, and were solvable just from the first few lessons, B.2.1-B.2.2. It is also minimally tested in terms of weight, and even if #5 was to be considered to be from this part of the syllabus, the percentage only goes up to 6-7%.
My guess is that we will typically have one or two questions from B.2.1-B.2.4, and that we will also sometimes see one or two questions from the rest of the syllabus content as covered in B.2.5-B.2.9. Focus most of your attention on B.2.1-B.2.4, but also plan to learn the ideas in B.2.5-B.2.9 if you have enough time, since I do believe that these sections will eventually be tested.
B.3 Markov Chains (Syllabus A.6, 2-10%)
Fall 2015: 4 problems, 9% of the exam
This section was represented on the high side of the percentage range, balancing the low representation of B.2.
The chosen problems test only discrete Markov Chains, and only touch on material in lessons B.3.2-B.3.4.
My feeling is that we will always see at least one problem from B.3.3, and often two as on this exam. I suspect we will usually see a problem from either B.3.4 as here or B.3.5 on branching processes. I also wouldn’t be surprised to see problems from B.3.1 that are more like the older problem types, even though these aren’t really the focus of the new source readings.
I don’t know what to say yet about the continuous Markov Chains, but I would expect them to be tested eventually, likely from lessons B.3.8 and B.3.10.
I expect that the material on Markov Chain Monte Carlo in B.3.6 and the material on continuous probabilities in B.3.9 to be comparatively rare, since it difficult to write good exam questions from this material. If I were pressed for time, I might watch the B.3.6 and B.3.9 lessons but not do anything else to prepare for the topics in those two lessons.
B.4 Life Contingent Cash Flows (Syllabus A.7, 2-8%)
Fall 2015: Two questions, 4.5%
tl;dr: Watch all lessons, ignore formulas except MII, Focus on B.4.1 and B.4.3.
Probably the most important observation to make here is that the two problems selected for the Fall exam directly test two of the three bulleted learning objectives in the syllabus section A.7. It would not surprise me very much if future exam problems are chosen to examine essentially the same aspects as the two problems that we see here.
I would suggest that you spend your time on lessons B.4.1-B.4.8 and just watch over B.4.9-B.4.11, and that you learn how to compute things but avoid memorizing the formulas other than the Most Important Identity in lesson B.4.5. I think that it is extremely likely that future exam problems will be able to be completed in a reasonable amount of time using the basic principles in B.4.1 and B.4.3, as long as you know enough of the vocabulary to set up the problems.
The computation needed for the first one in this section, #12, relies on the material in B.4.1 and B.4.3, but you still need to know what Term insurance is to answer the questions, and that’s in B.4.6.
In the second problem from this section, #13, terminology is used in the problem that does not appear in the study note. The phrase “equivalence principle” is not in the study note, although it should be clear what the “equivalence property” mentioned in the syllabus refers to based on the text of the note. Equivalence principle is the standard lingo from the old LC exam.
Section C: Statistics (Syllabus section B.1 – B.4, 20-40%)
17 problems 38%
I expect this material to be tested on the high side regularly, because I expect it to be difficult to write enough good GLM problems to test that section on the high side of its percentage range.
I found the exam questions in this section to be fairly routine. If you watch the lessons in section C and complete most of the exercises, you should do well on this section. The 17 problems that appeared came from 14 different lessons in the seminar, and no lesson had more than 2 problems.
The only thing I would consider watching without mastering is lesson C.3.6, on the Run test. This has never been tested, and is only in the seminar because there was an example problem released for exam ST on the topic.
Section D, GLM (Syllabus sections C.1 – C.4, 25-40%)
Fall 2015 exam: 11 problems, 25%
There is enough to say here that I already made an entire video about it. See lesson D.0.1, How to handle GLM Spring 2016.
Section E. Time Series (Syllabus section D.1-D.3, 5-10%)
Fall 2015 exam: Four questions, 9% of the exam.
Three of the four questions that appeared on the exam came from lessons E.2.2, E.2.3 and E.2.4. The other question is covered in E.1.4, where the cross covariance is defined. My guess is that this will be fairly typical. You’ll need to watch the E.1 lesson to get the background that you need to understand E.2., but most of the exam questions will probably come from E.2, and probably from E.2.2-E.2.4 in most cases.
I do expect them to examine Seasonal ARIMA sometimes, and I would expect to see a question from EWMA or Holt-Winters once in a while as well. I’m not convinced that they will include the Bass-Diffusion model on the exam, but it is in the readings listed on the syllabus. I’m also not sure there will be much emphasis on the R code aspects of the readings, since it is too easy to write good exam questions for this material without including those commands.