Many people have been wondering how Exam C compares to the new SOA exam STAM. They have about a 70% overlap, with the biggest changes between C and STAM being that Empirical Estimators (section C of the old C seminar) and simulation (section F) have been removed, and replaced with some new topics on health and property and casualty insurance. There are also some smaller changes, with things like Extreme Value Distribution (old lesson A.4.5), percentile matching (D.1.1) and method of moments (D.1.2) also being removed.

In terms of how the TIA C seminar compares to the STAM seminar, Sections A and B are fairly similar so if you have started those in the C seminar, there is no need to rewatch those lessons in the STAM seminar. The lesson that is most different in those sections is A.3.5, in which I’ve included a 2nd approach using raw moments to find the variances. After that, the STAM seminar has more differences from C. The removal of Empirical estimators means that Parametric estimators has been moved from Section D to C, and the old chapter D.1 has also been removed. The simpler cases of the method of moments are still useful in shortcuts for maximum likelihood estimators and are now discussed in the relevant pieces of the MLE chapter.

Lessons on the new material are still being completed. Currently the STAM seminar includes everything through Classical Credibility. Lessons on the remainder of credibility will be completed by the end of May, and lessons on the new insurance topics will be completed by the end of June.

In terms of the relative difficulty of the two exams, the new material is potentially broader in scope, but the sample problems that the SOA has released so far are easier than the questions that were given on the removed material. That does not mean that the actual exam will end up being easier, as the sample problems and exam questions aren’t always written by the same people. I will be writing problems that try to include potential wrinkles not yet in the SOA samples to help prepare you for whatever they can throw at you.

For CAS students, C and MAS-II are very different exams, and I will discuss that some other time.

I’m happy to announce that we recently released a significant update to our iOS video app, which we are now calling TIA – Learn. You can download it from the App Store, or simply update your existing TIA app through the App Store.

Updates to the new Learn app include:

– Download unlimited videos to your device for offline viewing
– Fully optimized for all iPhone and iPad models
– Stream videos over WiFi or cellular data with option to download videos over WiFi only
– Track your progress as you watch lesson videos
– See all downloaded (offline) videos in one place (like making a “playlist”)
– Update user interface, including more intuitive video controls for skipping forward/back
– Adjust video playback speed on a video-by-video basis and also set a default preference
– Quick access to the discussion forum for each lesson
– Your lesson progress is synced across your devices and the TIA website. If you stop watching a video on the website or app, you can simply resume in that same spot using the website or the app.
– Free samples included without a TIA registration

We also recently updated the Android version of the app to do the same syncing described above. This was a big feature request, and we are happy to deliver it to all platforms. We have more Android updates planned in the future.

Keep in mind the new Learn app is in addition to the newly created Flashcards app that we released in 2017 for iPhone and Android. We plan to continue enhancing both apps and appreciate all of the positive feedback we’ve gotten so far.

We are excited to announce our new Flashcards app is now available on the iOS App Store! We also put together a short video of the app’s features.

There are a few C cards loaded into the app already if you want to go ahead and check it out, and we expect to finish adding cards for C over the next month or so.  The new cards will be loaded into your app as soon as they are added to our system.

This version is only the beginning, and we have more features planned. The Android version is also extremely close and should be ready soon.

There is a minor change to the syllabus coming for the October 2016 sitting. The Akaike Information Criterion (AIC) is being added to the syllabus, which means an extra slide will be added to lesson D.5.7, the formula will be added to the Section D formula sheet, and I’ll have to make a couple of new practice problems. The total amount of this change is small (the AIC is somewhat similar to the SBC already in that lesson) and as it doesn’t apply to people taking the exam in June, my plan is to not release the changes to the seminar until after the June sitting is complete.

I’ve updated the solutions to the D.3 practice problems that involve finding the MLE and MoM estimators with zero-modified distributions. The previous solutions used somewhat ugly algebraic approaches, while the new solutions use the methods presented in the lecture which are cleaner and I think substantially faster than the approach used in the text.

I’ve also updated the formula sheets to include confidence intervals for VaR-hat in section F (the formula sheets are in the Important Tables and FAQ/Key Course Handouts chapter in the start of the seminar)

I’ve uploaded 2 more review videos to the Further Review Material/Common Questions chapter, both on examples of mixtures from all across the syllabus. The 2 videos are fairly parallel, with the first being about computing probabilities with mixtures, and the 2nd being about computing moments with mostly the same examples as the first video.

The motivation behind these 2 videos is that a lot of topics on the syllabus are different ways of thinking about mixtures of distributions, and my hope is by seeing a bunch of them side by side that the formulas will become more intuitive and less some flavor of black box to memorize. My next goal is to start working on a new set of review videos at the end of each chapter, ultimately to replace the older review videos from the previous version of the seminar (which are still worth watching if you are looking for review material before the new ones are made).

I’ve added a 2nd “Commonly Asked Questions” video about lognormal distributions in the review section (Section G). If you pay too close attention to the numbering, you will see that I skipped from G.1.1 to G.1.3 — I’m still working on a lesson to go in between. This is the second lesson in the seminar on lognormal distributions, the first being the overview in A.2.4. This lesson is meant not to tell you what the lognormal distribution is, but instead compare a bunch of different ways in which we have used the lognormal distribution, comparing side by side things like method of moments and MLEs, or Bayesian and Buhlmann, both of which are different for lognormals.

In terms of importance, this varies wildly between exams. Some people get no questions about lognormals and complain about having studied them, while others get 5 questions and complain about having not studied them. Hopefully this lesson will help you get more comfortable with them in case you end up in the second bucket.

I am working on a series of review videos designed to answer some commonly asked questions. Everything in these videos has been talked about in other places as well, but the point of these videos is to compare different cases side by side that a lot of people get confused about. For example, the first video that I posted yesterday is about the different denominators that we use. Sometimes we divide by n, sometimes n+1 (namely with smoothed empirical percentiles and p-p plots), sometimes we divide by n-1 (when looking for an unbiased estimator of the variance, which is used in empirical credibility and simulation other than bootstrapping).

The other videos in the series that I am currently planning are talking about mixtures, and talking about lognormal distributions (when do we use ln(X) and when do we use the moments from the tables). If you have other suggestions, let me know. I will eventually do a video about what ‘n’ to use in Bühlmann formulas, but that will probably go in a credibility specific review chapter.

If you are taking the exam in February, it is time to start taking practice exams. When you do that, you want to be sure that you are reviewing them carefully and learning from your mistakes. The reviewing process is both the most important part of a practice exam, and the one that people are most likely to overlook.

I’ve posted a video that talks about this, but the TLDW version is that the way that I would approach a practice exam is to spend 2, 3, or maybe even 4 days on it. I would take it one day, under strict exam conditions with no formula sheets, and find out what my raw score is. On day 2, I would redo the problems I missed with my formula sheet to see if I had any new ideas or if any formulas would help (if they would, then that is a formula I need to memorize!) Then I would review the solutions to all the questions I missed, as well as any questions that I got right but found hard.

Now comes the part that people skip: On day 3, I would redo all of the questions that I missed to help solidify the ideas I learned from the solutions. Passively watching a solution won’t help you remember something as well as redoing it, and this will give you your biggest improvement. If you can score 80% on the easiest 80% of the material, and guess on the rest, then you should pass, so you really need to learn from your mistakes and solidify that easy 80%.

Ok, but I said 4 days isn’t unreasonable. How does this stretch out that long? Maybe in doing these practice exams you find yourself missing multiple questions on core topics, such as MLEs or Kaplan-Meier estimators. If so, then redoing a bunch of practice problems on those specific topics can really help. Only do this with the most commonly tested topics — if you miss a question on ogive variances, or discretization, or some other really obscure thing, then this close to the actual exam the best strategy is to not care very much. That will be part of the 20% that you can guess on, and you certainly don’t want to waste a lot of precious, last minute time cramming in material that probably won’t be on the exam.

I’ve temporarily put back up some old review videos from a prior version of the seminar that a number of people have been asking for. The syllabus has changed slightly since these videos were made, so there are a few minor topics that are not covered in them (namely the non-inversion methods of simulating normals and (a, b 0) distributions, lessons F.1.4 and F.1.5, and the mortality table construction from C.3.3 – C.3.5). The review videos also cover Anderson-Darling, which is no longer on the syllabus.

Long term, the goal is to replace these with better videos. In particular, I want to make some lessons focusing on common confusions that people have and addressing ways in which the different topics are interconnected. Unfortunately I’m struggling a little bit with how exactly to do that.