Question

Consider the following model:

?i= B0 + B1D1i + B2 D2i + ui

where ?1i= (0 if person is nongraduate and 1 if person is graduate)
and D2i = (0 if person is graduate and 1 if person is non graduate)

and yi denotes the monthly salary.

(a) What is the problem with this model?

(b) How are you going to fix the problem? Suggest two ways to fix the problem.

Answer 1

(a ) The problem with this model, is

D1i and D2i are indicator variables of the same predictor variable - 'gradudation'.

Further, D1i = 1 - D2i. i.e. they are perfectly linearly dependent predictors which causes multicollinearity.

Further they are giving the same amount of information to model but unnecessesarily complicating model using more predictor variables.

(b) Two ways to fix the problem are:

1) D1i as explanatory variable.

2) use D2i as explanatory variable.