View Full Version : CAS 8 - Fall, 2012

Irish Blues
July 6th 2012, 01:14 PM
This is more or less a running summary of things to do so I know where I stand; it can also serve as an aid for anyone interested.

Lessons from 2011 [though this could apply to any upper-level CAS exam]:

1. Write. Write some more. Keep writing. Eventually, you will either hit the key words the CAS is looking for, or you'll so completely describe what's going on that if they fail to award full points, you have proof that the grading process is solely focused on "say exactly what we want" more than "show us what you know."

2. Work problems. Keep working problems, until you know them forwards and backwards. This means building examples in Excel, and generating as many random problems as you possibly can in as many different ways as you can possibly imagine. This is the only acceptable substitute for learning the numerical-based problems, because past exam questions only go so far.

3. If you only think you know it, you don't know it. You have to know the concept and how it applies before you can hit the "OK, I know it" mark that is necessary for the exam.

4. Don't assume that because there's a problem with given information that the CAS knows what it's doing or that it was accurately written. This sounds silly, but repeatedly [including at least once and arguably twice on the '11 exam] there are examples where the CAS clearly didn't check to see "did the data we gave really make any sense?" If there's a mistake, document it and explain why and state assumptions from there - and then be ready to write an appeal to explain why someone '''''ed up.

5. Did I say "write a ton?" Seriously, fill up the page if necessary; leave nothing to chance, because the CAS has shown it can't be trusted when it says "we look for reasons to give credit, not reasons to take it away."

Irish Blues
July 13th 2012, 02:42 PM
Finally found study materials; that's good, because I was afraid I was going to have to re-print everything. Pulled out notecards and re-sorted; tossed aside everything on Cummins [which is no longer on the syllabus]. Areas to especially focus on this time around:

1. Clark - botched these problems last time, and it's not going away.
2. Robertson - they'll probably get more in-depth here.
3. Couret-Venter - same thing.
4. Grossi-Kunreuther - last time's question was expected; I think they'll get much more abstract this time.
5. Mahler's papers on excess ratios and fitting curves - last time's problems were explanatory, I expect them to be much more like the original problems asked [computational] along with being abstract.
6. Gillam & Snader 3 - hammer away at the deductible sections, they always lob one out of here.
7. Skurnick - they've kind of ignored this for a while, they won't stay away from it forever.
8. Miccolis - because it's so damn easy to ''''' things up [like I did last year].

109 days to the exam. It's time to finally slay this monster.

Irish Blues
July 17th 2012, 01:45 PM
First 3 articles down: AAA, Bailey-Simon, Mahler 1 [on shifting risk parameters]. Also found my old Excel file with examples for computation problems, decided it's time to update this and fill it out completely. Bailey-Simon is complete [thought I had to build this from scratch, didn't get stuff that made sense at time; luckily I found something that I wrote a couple years ago that's beautifully written], though I could stand to put a couple more twists in just to be sure. Mahler 1 is done, simply because (A) there's only so much stuff that's really testable, and (B) for a few of the concepts, if you understand what's said you can figure out the answer in any numerical example.

Next up: Anderson. This is probably noted out about 85% of the way, but there's a couple of things I want to still jot down. The big thing here will be coming up with testable problems; I think I can do it, but it will take some work - so I've allocated about 2 days to figure that out.

Irish Blues
July 19th 2012, 09:46 AM
Couret-Venter and Robertson: I still think given the knowledge statements on the syllabus, there's only so much in each paper that's really testable to begin with - and even less that's testable in an exam setting. I pretty much nailed the questions that were going to be asked last year, so now I'm trying to figure out where they could go; Robertson, they could have us do a [simple] clustering example, and with the move to 4 hours I think that's entirely fair game. Couret-Venter, ... I need to think and ask around here.

Irish Blues
July 23rd 2012, 01:26 PM
I was thinking about going to a seminar this year, but I'm re-thinking that stance ... not because I don't think I'll get anything out of one, but because of $$$ - and not "gosh, the seminar is a lot of $$$" but more "holy, ... really, a hotel room is going to cost THAT much per night?" Yes, I know ... I'm not really paying that cost out-of-pocket, the employer is, but I still feel a duty [others don't] to be reasonably smart about what I spend even when on the company's dime. The seminar at ~ $800? OK, fine - that's the going cost, I'm ok with that. A hotel for $300+ per night for 5 nights? I'm really struggling with that part.

Irish Blues
August 2nd 2012, 07:09 AM
Thus far, with 89 days to go:

-- All of Section A has been re-written, with the exception of Couret-Venter and Robertson. I'll come back to these after making the first pass.
-- Miccolis, Skurnick, Lee, and the other two Mahler papers have been re-written. This includes examples in Excel for both of the Mahler papers which largely replicate the exhibits in each paper, but are not randomized. [Yet.]

Next up on the list:
-- NCCI 98, the ISO GL manual, the NCCI ER manual, and the NCCI RR manual. Mostly loose notes, but I'm going to work on an Excel example for at least the ISO GL manual.
-- Gillam-Snader. This includes Section 3 on the various types of deductibles.
-- Gillam, Venter, Teng, and Fisher. Mostly notes, either I have Excel examples already or I just need to quickly modify them.

That will hopefully get me to Clark in about a week and a half, and I'll probably spend a full week on it alone just making up notecards and Excel examples. I'm in good shape right now, as I'm not really expecting to be in serious crunch mode until 60 days out.

Irish Blues
August 4th 2012, 04:04 PM
OK, not NCCI/ISO stuff. I went to Gillam-Snader first. Went back through and re-wrote notecards, and found an example of the disappearing deductible in Excel that I got a hold of some time back. Doesn't really make any more sense, but at least I have something to look at when studying. Stuff in Gillam-Snader 2 [Table M, balance equations, etc.] makes a little more sense. If nothing else, it's got me thinking about ways some equation can be used and how to quickly manipulate it to solve for a specific variable.

Trying to decide if it's worth taking notes on the ex-medical discount section at the end of Gillam-Snader 3. I think there's been a question on this in the past, but that was like 10-12 years ago; it's fair game and there's a decent argument for seeing it show up but I haven't decided if it's really worth taking time to commit to memory [u]and work examples on.

Irish Blues
August 9th 2012, 01:30 PM
Done through Clark and Bernegger. I'm taking the same approach to each - there's a bunch of theoretical stuff on both papers, but the chances of any of it showing up on the exam is near zero [though I may scan old exams to see if any of it from Clark found its way into an exam question].

Now off to Grossi-Kunreuther [the CAT modeling paper], then off to the NCCI and ISO manuals. On pace to be largely done with notecards by the 18th, which still leaves time to put together anything that jumps out but allows me to turn around and focus on putting together numerical problems.

Irish Blues
August 17th 2012, 03:12 PM
Close to being done with notecards; Excel examples will be the next task.

One thing to chew on, though:
-- Venter says that "for excess lossees, by eliminating the first small portion, enough losses are eliminated to bring up the average value and to reduce the probability of a loss being a large multiple of the average value. This makes the excess losses less heavy-tailed and thus more predictable than total losses."

-- Teng says in explaining why an LDD plan is riskier than a fully insured plan, "excess loss is harder to estimate than total loss."

Are the two statements inconsistent with each other?