Behavioural modelling - proofing the selections

The FastStats Modelling Environment is a powerful tool; through its simple interface you can create very sophisticated selections. Once you have your analysis and base selections, you may want to proof them before going on to analyse them.

You can view the created selections from two places within the modelling environment.

  1. Right click on the selection in the selections tab and choose View a Copy

Or:

  1. From within the selection builder window, select View a Copy.

Base group

Alongside the prerequisite clause, the base selection has one clause identifying the bookings that took place in the 6 month period before the reference date. The reference date is the training date, in this case 01/01/2022.

Analysis group

Alongside the prerequisite clause, there are two further clauses. To be a member of the analysis group, a person must also be in the base group. Hence, the middle clause is the same as that in the base selection, and the final clause identifies those people who took out insurance cover in the 3 months after their booking.

The Data Grid tool is an ideal way to sense check your selections and you can access it directly from the modelling environment.

  1. Click on the Data Grid icon.

  2. Click Build

The data grid is pre-populated with appropriate columns. For each person, the columns identify the date for Event 1 and, where applicable, Event 2. A third column confirms if each person is in the analysis group or not.

People in the base group will only display Event 1, whilst those in the analysis group will also display Event 2.

To meet the criteria for both base and analysis, a person's Event 1 (their booking) must have taken place up to 6 months before the defined training date - here 01/01/2020. As you’d expect this is the case for every person. For example:

  • Person URN 18852 is in the base selection and we can see that they made a booking in July - i.e. approximately 5 months before the training date. However, this person did not go on to take out an insurance policy in the next 3 months and this is reinforced by "No" in the Analysis column.

Person URN 763270 is in the analysis selection, having made a booking in December and then taking an insurance policy out just 4 days later. Here the Analysis column displays "Yes".

 

It is possible for the policy to be taken out after the training date - see below.

In this scenario, where the training date is 01/01/2020, a person with a December booking date could purchase their policy after the training date - i.e. in January, February or March.

 

For related topics, see Behavioural modelling - setting up the environment