There are 2 questions in picture.

Suppose savings is explained by the following MLR model: sav = Bo + BIinc + Bage + Bymale + u Suppose you also know that E(u inc, age, male) = 0 and in addition that Var(u | inc, age, male) = 02 inc2 a) Transform the model so that the error term is homoskedastic (write the new model down). b) Demonstrate that Var(u | inc, age, male) = o' where u is the error term of your transformed model.Consider the following sample regression function explaining term GPA for students in fall and second semesters: trgpa = -2.12 + .900 cragpa + .193 cumgpa + .0014 tothes (.55) (.175) (.064) (.0012) [.55] [.166] [.074] [.0012] + .0018 sat - .0039 hsperc + .351 female - .157 season (.0002) (.0018) (.085) (.098) [.0002] [.0019] [.079] [.080] n = 269, R2 = .465. Note: cragpa is the GPA average (overall among students) for courses the student has taken, and cumgpa is the student's own cumulative GPA prior to the current semester a) Explain whether or not the signs of Persona and Scumgpa make sense. Test the significance of each of these coefficients at the 5% level. Does it make a difference whether you use heteroskedastic-robust standard errors? b) Test the significance of Brothers (coefficient on total credit hours) at the 5% level. Does it make a difference which standard error you use