The summary sheet previously showed that the expected value of X_p was conditional on X_L being greater than d.  This conditioning should have been either X_L >0 or X>d.  I’ve corrected it to say X_L>0.  This also affected one flashcard from C.3.6.  I have that corrected in the flashcard app.  Please change it by hand if you have already printed out printable flashcards.

The three new lessons C.3.4, C.3.5 and C.3.6, the summary sheet for C.3, problem sets, and flashcards for C.3 are now complete and available in the course.

(This new content was announced first here: http://theinfiniteactuary.com/blogs/examS/2019/03/12/c-3-probability-for-insurance/)

At the bottom of the Model Estimation section, when estimating the response means on the second iteration and after, the link function should have been included.

This has been corrected.  Thanks to Marco for pointing this out.

At some point, the CAS added several new topics to the MAS-I syllabus.  This came to my attention last week, so there will be a few new lessons coming out this week to address these new topics.

They all fit within the context of the subject “Probability for Insurance”, so I’ve decided to rename C.3 from “Order Statistics” to “Probability for Insurance”.  This lead to C.3.1 -C.3.3 being updated to show this new chapter title, but that’s the only change that has occurred for those three lessons.

The first new lesson just posted, on Transformations of random variables.  There will be two more lessons, one on Tail weights of distributions, and another dealing with policy modifications like deductibles.  I hope to record these over the next few days, and I will also be adding exercises and flashcards to go with these lessons soon.

Thanks to Nicholas and Ali, both of whom pointed out this error in the lesson within an hour of one another this morning.

For n observations when the number of variables is n-1 the number of parameters will be the maximum allowed, which is n-1 + 1= n.

So, if we have a problem because d = n-1 is the maximum number of variables we can handle.

So for the forward selection, we must stop at if

Also, on backward stepwise selection, if then we cannot start the process, it will only work when

The confusion here typically stems from d being the number of variables and not the number of parameters total, which would be d+1.

I have this fixed now, in the D.5 summary sheet and the D.5.1 lesson video and handouts. And the flashcards.