A lot of bookwork questions, and nothing seemed particularly complicated. I mucked up a few of them but wasn't expecting to do very well anyway.
hmmm.... Is it just me who found that paper impossible?!? I thought that was by far the the hardest CT6 paper over the last few years! I studied really really hard for that exam and was very confident going in. I got over 80% in my mocks (Apr 08 & Sept 08) but found that exam impossible! Please tell me I'm not the only one??
Overall I found the paper to be OK. I was stumped by the minimax question though - None of the options seemed to be dominated? Was annoyed I couldn't get the Gamma as an Exponential question out under exam pressure >< Time Series - took a stab but not really confident there. Rest of the paper was quite decent I thought. I wonder what kind of raw mark you would need to pass this subject?
Here's my notes on the questions and how the marks were divided up: 1. Liability insurance - principle characteristic & 4 distinct types [4] 2.Gamma GLM - natural parameter and scale parameter [3], canonincal link [1] 3.Bayes - Beta posterior for theta - 3 policies, 9 claims [2], derive estimator under all or nothing loss [4] 4. Decision theory - complete table [5], any dominated? [2], minimax [1]. i thought no insurance at all was dominated by proportional insurance 5. Sum of 3 compound poissons, individ claim distb exp (1/4); prove E=E[E(S/lambda)] [2], show E[S/lambda]=4lambda and var[S/lambda]=32lambda [4], calc E [2], calc var [2]. 6. Run off triangle ACPC [10] 7.Monte Carlo acceptance-rejection - derive algorithm [5], proportion accepted [1], is it more efficient to use uniform? [4] 8. Ruin theory - explain what U(t) is [2], explain phi(U,t) and phi sub h (U,t) [2], compare 3 pairs [6] 9.XOL insurance, M=1+k - find E[X] and var[X] and simplify expressions [5], find p(S>700) given lambda=500 and N distb Poi(lambda) 10. Time series - show not stationary [3], show differenced X is stationary using autocovariance function of Y [7], show differenced process is invertible and bigger variance than Y [5] 11. NCD criminal offence question - min thresholds [4], p(claim) [5], stationary proportions [7] I thought it was a fair paper - topics as expected given syllabus and nothing deeply horribly unattemptable. But i was seriously pressed for time and not sure i managed to do enough to scrape through. Best of luck to all
