Unformatted text preview: T_ 0
Std Error 0.5756814265
6.6543284337 Sum of Squares 6.9645989789
Prob > F
0.2175034266 9 5) MULTIPLE LINEAR REGRESSION
We consider automobile injury claims data. The data contain information on demographic
information about the claimant, attorney involvement, and economic loss, among other variables.
We consider here a sample of n = 1,340 losses. More specifically, we consider the following list of
LOSS Whether the claimant is represented by an attorney (=0 if no and =1 if yes)
Claimant’s sex (=0 if male and =1 if female)
Claimant’s marital status (married, single, widowed, or divorced/separated)
Whether the claimant was wearing a seatbelt/child restraint (=0 if no, =1 if yes)
Whether the driver of the claimant’s vehicle was uninsured (=0 if no, =1 if yes)
The claimant’s total economic loss (in thousands) Consider the attached JMP output for the regression model of the logarithm of the losses (LN_LOSS)
on all the variables above plus:
• CLMAGE_sqr è༎ the square of the claimant’s age, • CLMAGE*ATT è༎ the interaction between claimant’s age and the attorney dummy
variable, è༎ • CLMAGE_sqr*ATT the interaction between the square of claimant’s age and the
attorney dummy variable. Moreover, the marital status has been split in the dummy variables MARITAL_SINGLE,
MARITAL_WIDOWED and MARITAL_DIVORCED. a) Provide an assessment of the attached regression model.
b) Provide an interpretation for the CLMAGE*ATT coefficient.
c) What was the objective of the analyst who did run this analysis in using so many terms
involving the claimant’s CLMAGE_sqr*ATT)? age (i.e. CLMAGE, CLMAGE*ATT, CLMAGE_sqr, Do you think the analyst’s idea is supported by the data? Why? d) Do you have any suggestion on how to improve the model? 10 and Response LOG_LOSS
Whole Model LOG_LOSS Actual Actual by Predicted Plot
View Full Document