ICRFS-Plus™ Training videos for new users
If for any reason you are unable to view the training or demonstration videos, please contact our support staff at support@insureware.com and we will arrange to send you a copy of the videos on CD-Rom. You will be able to run the videos from the CD.
The training videos should be used for hands on training. We suggest you run the videos on a separate computer using a data projector, and train as a group.
The only way you will learn all the new concepts and be able to exploit all the immense benefits is by using the system. Experiential learning is imperative.
It is important that you study the videos in sequential order as set out below.
Table of Contents
1. Introduction to ICRFS-Plus™ 10.2 and modelling modules
2. Modelling using the Link Ratio Techniques and Extended Link Ratio Family modules
3. Introduction to the Probabilistic Trend Family modelling framework
4. Modelling real data (CTP) in the PTF modelling framework
8. Further PTF Modelling Examples
10. Uneven Sampling periods, updating, and various other topics
- 10.1 Uneven Sampling Periods and collapsing
- 10.2 Accident Year Hetero, Model Parameters, and the testing the Wizard
10. Uneven Sampling periods, updating, and various other topics
10.1 Uneven Sampling Periods and collapsing
What does an uneven triangle look like? We explore "C uneven" an annual by quarterly dataset from the Workbook and compare it with "C even" a quarterly by quarterly representation of exactly the same data, made by interpolating zero rows into the array. "C even" represents the way that we would have gained access to this data for modelling purposes prior to version 10.2.
We show that modelling with the two versions produces identical results. This is immediately true of the SGI starting model, but for the wizard to produce identical models in either case we must set up the modelling preferences for the wizard to respond identically when there are few data points. We show how to do this and then how to save and load customised wizards.
The forecasts for the two cases are also seen to be identical, but an important change in the forecast table is shown, since the summary by calendar periods can no longer be placed next to the first development year column as they will be over very different sizes.
We consider the issue of smoothing a model after the wizard has placed a large number of parameters. This is more likely to arise with uneven sampling periods in the calendar or development directions as they may have a great many periods.
How to remove parameters. How to detect seasonality and where present to interpolate seasonal effects into the forecast scenario.
We illustrate collapsing by collapsing "C even" to create a replica of "C uneven". The need to match the starting period is stressed.
Updating uneven triangles by the addition of diagonals works in the same way as updating even triangles. We show how previous models are carried over to the updated TG and how this makes for ease of monitoring of forecasts and targets.
The video demonstrating Uneven Sampling Periods, Collapsing and Updating is approximately 35 minutes long.
10.2 Accident Year Hetero, Model Parameters, and the testing the Wizard
The contents of this video include more on Accident year hetero, how to store model parameters and how to test the wizard.
We look at how to test for accident year hetero in real data. Two TGs from the Workbook, ABC and Wcom are analysed. We begin with a wizard model and then choose Manual Hetero/Accident year hetero. A straighforward procedure of fitting all variances and then optimising suffices to test for significant accident year hetero changes. In both of these cases accident year hetero is ruled out. Accident year hetero usually indicates a change in the mix of risks and hence there are usually other trend changes involved, so MPTF modelling is recommended.
When we save a model in ICRFS-Plus we are actually saving a parameter structure. The working parameters are recalculated from the data each time we open a model, and this is what makes it interesting to transfer models between different TGs. It is also possible to save the exact parameter values as well as the structure. In this case parameter uncertainty is no longer a factor. We do this via Test/Save Model Parameters. We show how to load such a parameter model and note that this kind of model does not include any zero-weighting of outliers such as might have been a part of the original model. We can obtain exactly the same forecast output from a normal model and a parameter model if we exclude parameter uncertainty in the forecast from the normal model.
We note that parameter uncertainty contributes not only to standard deviation of the forecast values but also increases the forecast means.
We create a simulation from a set of parameters that has almost no process variability. In this case the wizard exactly recreates the original set of parameters. As we increase the process variability the wizard estimates are likely to deviate more and more from the values used to create the simulation. This is because trends can be masked by process variability. Awareness of this is important in formulating good forecast scenarios. For example if the final calendar trend in a model is zero this may be because an underlying positive trend was picked up but had too much associated variability to survive optimisation. We can place parameters to inspect these trends in order to create conservative forecast scenarios.
The video demonstrating Accident Year Hetero, Model Parameters, and the testing the Wizard is 15 minutes long.
