Hi Allstat
Quick question on risk analysis.
Say it is to do with car loans.
At a macro level, there are three categories each with three levels
(They do not day how they calculated these "risks", I assume it is the historical percentage of defaulters)
Length of Loan
1-2 0.25
2-4 0.35
4+ 0.50
Car
Low Range 0.2
Middle Range 0.1
High Range 0.05
Company
A 0.10
B 0.12
C 0.15
The question of interest is "what is the relative weights of the categories" on an individual basis.
Now if we have individual data on each person, so for example Length2-4, Car Type Middle Range and with Company C is it correct to multiply the risks associated with each category and then perform a regression analysis on this response variable and use the standardised beta co-efficients as indicators of the relative weight of each category ?
(This is what the owner of the data proposes).
Cheers
Kieran
|