2. (15 points) This preliminary exercise prepares you to perform and visualize confidence set estimation using MLE. For concreteness we focus on the case d=2. We are interested in visualizing (when d=2 ) the ellipsoidal set NC(c,):={y=(y1,,yd)Rd:(yc)C(yc)2} where C=(cij) be a symmetric (strictly) positive definite dd matrix, cRd and >0. The boundary NC(c,) of NC(c,) is NC(c,)={y=(y1,,yd)Rd:(yc)C(yc)=2} (a) Suppose C=[1002] is a diagonal matrix, where 1,2>0. Find a,b>0 such that NC(c,)={y=(y1,y2)R2:a2(y1c1)2+b2(y2c2)2=1}, which is an ellipse in "standard orientation". In R, draw NC(c,) where c=(1,2),C=[2005] and =0.5. (It is ok to plot only the boundary which is an ellipse.) Hint: Start with the trajectory of x=acost,y=bsint,0t2. b) Consider now the general case where C is strictly positive definite and admits the diagonalization C=SS where S is an orthonormal matrix. Write in R a function add_boundary (c, C, epsilon) which adds to an existing plot the boundary of NC(c,). Use this function to draw the boundary of NC(c,) where c=(1,2) and C=[2115]. This function will be used in the next problem. Hint: yCy=ySSy=zz where z=Sy. In R the eigenvalue decomposition is implemented by the function eigen(). You can also check your plot against the output of WolframAlpha: Try e.g., https://www. wolframalpha.com/input?i= plot +5x%5E2+%2B+2xy+%2B+10y%5E2+%3C%3D+1.