Consider the following linear regression model for household food expenditure: FoodSpend=0+1Income+2Age+3HHsize+(C1) where: FoodSpend is household expenditure on food (in dollars) Income is the household income (in $1000 s) Age is the age of the identified household 'head' (in years) HH size is the number people in the household Assume that E( Income, Age, HH size )=0. Using the information above, answer the following 4 questions: (i) The error variance in model (C1) might not be independent of the regressors. Outline the steps you need to follow to undertake the Modified White (MW) test for heteroskedasiticty. Be sure to state the null and alternative hypotheses, the test statistic and the conclusion you would make from rejecting the null hypothesis. [2 marks] (ii) From conducting the MW test in part (i), you find that there is evidence for heteroskedasticity. What are the implications for your coefficient estimates if you use OLS to estimate model (C1)? What about the coefficient standard errors? [ 2 marks] (iii) Suppose you know that the heteroskedasticity detected in model (C1) takes the following form: Var( Income, Age, HHsize )=2 HHsize Write a transformation of model (C1) that has a homoskedastic error term. [3 marks] (iv) Suppose instead that you do not know the functional form of the heteroskedasticity in model (C1), Explain what alternative approach you would take in estimating the model. [2 marks]