This is a monte Carlo question and needs to be solved algebraically or via EViews or both ways.

Heteroskedasticity:

Suppose our true model is yi = βo+β1*x1i+ β2*x2i+ e(exp(x2i))^0.5 , i = 1, 2, ..., N, et ~i.i.d.N(0, σ^2). where i = 1,2, ...,N. Compare the estimators and test statistics when you take heteroskedasticity into account to those you obtain when you ignore her- heteroskedasticity. Your investigation should include consideration of different sample sizes. An outstanding report should try to address the following issues with both algebraic proof and experiment results.

• Does heteroskedasticity in this topic affects the t-test validity for both B1 and B2 if one forgets to take account of heteroskedasticity when calculating the t-test statistics? Why?

• How many tests can be used to detect the presence of heteroskedasticity?