I am part of an actuarial working party.
it is called the bayesian methods working party.
the chair is Saunak.
I am a member. it has been in existence for 18 months.
12 months ago he encouraged me to join him in submitting a synopsis for a presentation at a conference. i had barely started my analysis but took a risk gave him something
in the end nothing was arranged for us by the conference, i guess they chose other submissions. it was a relief given i wasnt prepared
recently an opportunity for data science conference organised by actuarial profession has come up and Saunak has resubmitted the same thing
i am not disappointed per se with him submitting because had a heads up - just for the record.
i will give you more details.
Here is the synopsis:
Although Bayesian Techniques are very popular in academia and a lot of research has been done on Bayesian applications in insurance, bayesian techniques do not get their due creidt in the insurance industry.
The bayesian working party wants to showcase some interesting applications of bayesian techniques which can be used in both Life 7 P&C insurance.
Topic 1: bayesian principal component analysis of interest rates (this is my topic)
the principal component analysis is extended to use a bayesian framework. we show how incorporating expert-informed priors can improve forecasts when changes in the economic environment are not yet reflected in historical day.
in particular we consider:
* why a classical model would have overestimated interest rates during the prolonged low-rate period foloowing the 2008 financial period.
* how a bayesian model could have adjusted expectations more appropriately
* how similar adjustments can be applied in response to global tariff policies imposed by the US administration.
Topic 2: claim reserving - projecting loss development factors using (hierarchial) non linear generalised additive models (this is saunaks topic)
while traditional actuarial reserving techniques works well for lines with credible claims it will not work for excess portfolio which are generally for higher layers where claims data is sparse.
while traditional actuarial reserving techniques works well for lines with credible claims, it will not work for excess portfolio which are generally for higher layers where claims data is sparse. in particular:
* we can showcase how bayesian techniques work
* we can compare the loss development factors & ULRs using traditional vs bayesian techniques
in this session we will have high level presentation & will be delivered virtually featruring variaous members of the bayesian WP. this session should be of interest to actuaries with interest in data science.
the problem is ... the analysis is still not complete and i fear getting submission accepted
saunak says submission likely to be accepted because this time it is the overarching body of which our working party is a part that is organising the conference.
he really wants us to make a submission.
i have the following concerns:
* i need to change direction of my analysis
* i dont have technical knowledge yet to do the analysis
* i dont have time to do the analysis.
he is going to give me timelines for presentation which then goes for review.
im meeting him tomorrow for lunch.
i want to help and present something...
Sunak may suggest... just outline what you are investigating...
if i go with lighter pice of commentary.. what should it be on
weak or incomplete work work gets noticed
i am unsure about it
i am confident presenting but...
the synopsis doesnt reflect where my work is right now...
concerned about get agreement on submission and figure it out later....
I am concerned about delivering something to the required standard in this time frame
do i say im happy to help with presenting...
i need to go away and think about what you have said
i dont want to present something that is properly backed by analysis
a vague talk with no substance - could be judged poorly