(2012.Fall 2 b.) → 1.000 pts
I am confused with this problem and the answer below. Does missing data here mean one field is blank, for example we don't know if there is prior auto policy but know if there is prior HO policy? If that's the case, how do we use this data in the model to lead to the extrinsic aliasing etc?
"Missing data can lead to extrinsic aliasing. This occurs when there are linear dependencies in the observed data because of the nature of the data. In this case, the “missing” level for “prior auto policy” will be perfectly correlated with the “missing” level for “homeowners policy”."
Comments
This is a good example of the CAS making an easy question challenging through burying the important details in words.
We're given claim frequencies for "bodily injury liability coverage" which implies the insured has a current auto policy with the company. We're then told the data is grouped according to whether the insured had a prior auto policy or a homeowners policy.
The CAS model solution 2 assumes there are some current insureds which lack information about both their prior auto and homeowners policies. In which case these are perfectly correlated so we have aliasing. It could be possible to be missing information for only one of the prior auto or homeowners fields though which could introduce multicolinearity. If in doubt, make sure to state your assumptions in the exam.
As a side note: The definitions of intrinsic and extrinsic aliasing are no longer on the syllabus. Instead, make sure you understand multicolinearity and aliasing in general.