A5-A10 - Meyers (Monograph #1) Flashcards
According to Meyers, what is the criteria required for “general applicability of the model”?
Percentiles of the outcomes of a triangle should be uniformly distributed.
What shapes of histogram and p-p plot implies that the distribution of the modelled results are uniform?
Flat histogram, and p-p plot distributed on a 45 degree line.
What shapes of histogram and p-p plot implies that the distribution of the modelled results are too light in the tail of the triangle?
Histogram int he shape of a “U”, and when the p-p plot has a shape like an “S”.
What shapes of histogram and p-p plot implies that the distribution of the modelled results are to heavy in the tail of the triangle?
Histogram in the shape of a dome, and when the p-p plot has a shape like a “~” (tilde), or shaped like a backwards “S”.
For the Mack Model, for incurred claim triangles, what conclusion can we draw from the p-p plots?
The Mack model’s estimated results are too light in the tail compared to actual. (They all have p-p plots that are “S” shaped)
For the Mack Model and the Bootstrap ODP model, for their respective paid loss triangles, what conclusion can we draw from the p-p plots?
That these models are biased in such a way that the resulting estimates are higher than actuals.
What are the 2 Bayesian MCMC models mentioned in the paper?
- Leveled Chain-Ladder (LCL) Model
- Correlated Chain-Ladder (CCL) Model
What improvements, if any, did the 2 Bayesian MCMC models have over the Mack Model?
The Bayesian MCMC models still had patterns to their p-p plots that implied that there estimates are light in the tail, but overall were an improvement in terms of fit (Satisfied the Kolmogorov-Smirnov Test more often).
What are antoher way to describe each of the terms in the following formula:
Var[X] = E[Var[X|Θ]] + Var[E[X|Θ]]
Total Risk = Process Risk + Parameter Risk
