Q6 (ii), Sep. 2015, ST7

Discussion in 'SP7' started by Smith, Jul 4, 2021.

  1. Smith

    Smith Very Active Member

    as per the solutions of the examiners' report, i have a couple of thinking, to consult, as below.

    1) the question is related to the range of reasonable best estimates, but from my understanding, it's the approach introduced in the part 3, estimating the range of possible outcomes, Chapter 18, and then, in the latter part of the solutions, it stated "...... exclude ...... events or claims with very low frequency and high severity, as the definition of reasonable might exclude such extreme occurrences". so my question is that, may i understand that the three approaches mentioned in the textbook, i.e. stochastic models, alternative sets of assumptions and scenario testing, can be used for derive the range of possible outcomes, reasonable outcomes and best estimates? e.g. by excluding the extreme data to derive the range of reasonable outcomes from that of possible outcomes, and furthermore, excluding those data that despite reasonable but not the estimate of mean or average outcome?

    2) in the 3rd paragraph of the examiners' report, it stated "The bootstrap model includes parameter and process error but makes no allowance for model error". but in page 10, Chapter 18, of the textbook, it sated "(the range of reasonable outcomes) ...... allowing for parameter / model uncertainty and some element of process uncertainty as well", seems contradict each other. meanwhile, in page 12, Chapter 18, it stated "these models ...... attempt to quantify process and parameter error ......". would you kindly advise, is there a place having a typo? because seems not aligned each other.

    thanks!
     
  2. Ian Senator

    Ian Senator ActEd Tutor Staff Member

    1) Sorry, you've lost me, can you please simplify your question? Or split your questions out if there's more than one there? I think what you're getting at is the interpretation of 'a reasonable range', which is what the examiners wanted you to comment on. We discuss this further in ASET, eg you could argue that the choice of 25th and 75th percentiles do not necessarily represent a reasonable range. Hope that helps.

    2) A range of reasonable best estimates is what a number of actuaries might give you as their best estimate using acceptable methods. So this would include model and parameter error, but not process error. Bootstrapping uses the same model lots of times, so it would capture parameter and process error, but not model error. You could potentially adjust your bootstrap model to quantify model error in some way, but we don't know whether that has been done for this question. Again, all covered further in ASET.
     

Share This Page