Hardest parts for Exam 4: a survey

[Jim Daniel]

I'm helping an employer design some external assistance for their student actuaries preparing for Joint Exam 4.They're considering 3 or 4 days of lectures to help people get started with their studies.

I have two questions regarding how to go about this.You can reply here or to me at jimdaniel@mail.utexas.edu . Be as broad or detailed as you wish in responding.

QUESTION #1: Do you think it's more helpful to have 3 or 4 days of lecture on: A) a brief introduction to the main ideas on _every_ topic, or B) a full treatment of the hardest material to learn on your own from the books and various study manuals and study notes?

QUESTION #2: Here's a rough breakdown of the broad topics on the Joint Exam 4 syllabus. Which one or ones are the most difficult to learn on your own from the books and various study manuals and study notes?

1. Regression
1a. Parameter estimates and their variances
1b. Hypothesis tests
1c. Correlation, Durbin-Watson

2. Time series
2a. Smoothing & extrapolation
2b. time series & their properties (MA, AR, ARIMA, autocorr)
2c. Parameter estimation
2d. Forecasting

3.Credibility
3a. Full credibility
3b. Buhlmann, Straub
3c. Bayes
3d. (Semi-)emperical

4. Loss models
4a. emperical estimation
4b. Max likelihood
4c. Other parameter estimation methods
4d. Goodness-of-fit tests

5. Simulation
5a. Basics
5b. Bootstrap
5c. 2-sample & multi-sample tests
5d. Other goodness-of-fit tests

6. Survival models
6a. S & H estimators & their variances (Product-Limit, Nelson-Aalen, etc.)
6b. Hypothesis tests
6c. Kernel smoothing
6d.Proportional hazards & accelerated failure time models
6e. Other [likely to disappear from syllabus for May 2003]

Thanks.

Jim Daniel
jimdaniel@mail.utexas.edu

[Pi Man]

1. anything and everything helps
2. if i haven't started studying, how would i know what's going to be hardest to learn??

[burton leon reynolds]

1. anything and everything helps
2. if i haven't started studying, how would i know what's going to be hardest to learn??

These questions where obviously meant to be answered by someone who can add value to the discussion. Seems strange to me that this needs to be pointed out.

[M.]

I think it would be more valuable to teach the toughest topics. Teaching the easier stuff would seem a less efficient use of the time.

The hardest topics, IMO, are the non-parametric credibility problems and all the survival analysis stuff.

[Jim Daniel]

Sorry, I should have mentioned that these lectures would be about 12 weeks before the exam. No prior study by the students is assumed.

Jim Daniel

[Pi Man]

1. anything and everything helps
2. if i haven't started studying, how would i know what's going to be hardest to learn??

These questions where obviously meant to be answered by someone who can add value to the discussion. Seems strange to me that this needs to be pointed out.

brutal, man... get some manners, would you?

[Agtuary]

I agree with M. An overview of the topics is virtually worthless. If the material is easy to learn from the books I can get the same understanding from very little study on my own. If the material is difficult, an overview probably isn't enough to help me understand the details.

I can't help on what material is most difficult because I haven't started studying yet.

[glenn]

Actually, given that it's three months before the exam, I would personally prefer an overview of the entire course. The one thing that seminars do for me is provide me with the big picture. Having a good overview of what I need to know and a brief introduction to get me started would be what I would want. The tougher material I can review on my own later.

Just my 2 cents.

glenn

[phdmom]

I would prefer an overview of all topics. I wouldn't be comfortable with someone telling me "These are the parts we think are hard, the rest you can learn on your own." Especially if you don't expect the students to have started studying yet.

The topics I found hardest to learn were time series (MA, AR, ARMA) and survival models (esp proportial hazards).

[phdmom]

I would prefer an overview of all topics. I wouldn't be comfortable with someone telling me "These are the parts we think are hard, the rest you can learn on your own." Especially if you don't expect the students to have started studying yet.

The topics I found hardest to learn were time series (MA, AR, ARMA) and survival models (esp proportial hazards).

[jerkweed]

Jim, Having just sat for the exam, and given that the students have 12 ( or perhaps you could allow more ) weeks to learn the material in depth, I recomend the following:

As a prerequisie, The students must review their basic stats.

I would have an overview which covers the concepts on each topic 16 or 17 weeks before they write the exam, then self study for 12 weeks or so, then hold a review/ Q&A session for a few days or provide e-mail tutoring support from the time the seminar ends untill exam time. This way they would have an easier time comprehending the concepts each section is trying to convey. Ramping up to exam proficiency with unaided self study for C4 in 18 weeks is quite difficult.

Truthfully, I dont think all the basics can be conveyed in 3-4 days. 6-7 would be more ideal.

[Ac2ary]

1. If the purpose of the training is to be a first step in preparation for the exam an overview would be much more helpful. However, if a students preparation begins 12 weeks before the exam, they have a very limited chance of passing Course 4. If the students have studied 50 or more hours on their own, a study of the harder subjects/type of problems would be better.

2. There is no one topic that stands out as more difficult than the rest. The difficulty in passing comes from the volume of material and from the ability of the exam committees to create difficult questions out of relatively simple concepts. So where have the more difficult questions come from? Survival Models, Nonparametric credibility, continious Bayes analysis, theoretical Regression problems ("which of the following is correct?")

[4sigma]

For your situation where you're giving 3-4 days of lectures 12 weeks before the exam, I expect it would be helpful to include some sort of overview. You could probably keep this to a day or so, and use the remaining time to go in-depth on a difficult topic or two.
In terms of good topics to concentrate on, I would vote for the survival analysis. There are usually several exam questions on this and the available materials on this are not as easy to study from as the materials for most of the remaining major topics.

[M.]

I think the 'weed may be on to the best idea, though I don't know if it would be feasible. I agree that the students should be expected to do 50 or more hours stats review ahead of time for an overview class to have much benefit. Following up a few weeks before the exam with an in-depth class on tough concepts could be very helpful.

[jerkweed]

I know M. it won't be easy. Ideally, one should have been exposed to all the material long before the exam. I was just trying to guide the uninitiated to the path of least resitance. There is no silver bullet. Lets face it, the SOA is testing competency on 4 or 5 semesters' worth of material and there isn't an easy way to assimilate that much material in such a short time. :-?