(III) The output from the SAS programs in Part II.
In Part I, you can refer to plots or tables or large matrices that
problems ask for by saying (for example), ``The scatterplot or matrix for
Problem 3 is on page 17 of the SAS output.'' If necessary, add
page numbers to the SAS output, so that (for example) you don't have
several different page 1s in Part III.
1. Consider the matrix
A = ( 3 2 )
( 2 6 )
(i) Find two linearly-independent eigenvectors v_1,v_2 in R^2 and
their eigenvalues d_1,d_2. (Hint: One way to start is to solve
det(A-zI)=0 and then find vectors v such that (A-zI)v=0.)
(ii) Find an orthogonal matrix O and diagonal matrix D such that
A = ODO'. Verify that O an is orthogonal matrix, and verify that
ODO'=A. (Hint: This is the spectral theorem for A. Try
O=(v_1 v_2). )
(iii) Use part (ii) to find a symmetric commuting square root B of A.
That is, find a 2 x 2 matrix B such that B=B', AB=BA, and B^2=A.
2. Consider the matrix
A = ( 1 2 3 )
( 2 5 7 )
( 3 7 12 )
Find an upper-triangular matrix B such that A=B'B, and verify that
A=B'B. (Hint: See pages 25-26 in the text.)
3. (i) Let X_1 and X_2 be real-valued random variables. Show
that
Cov(X_1-X_2,X_1+X_2) = Var(X_1) - Var(X_2)
(ii) Let A=A' be a 2 x 2 symmetric matrix with
tr(A)>0 and det(A)>0. Prove that A is positive definite.
(Hint: Use the spectral decomposition of A.)
4. Let X = (X1 X2 X3)' be a
random vector in R3 with mean vector E(X)=0 and covariance
matrix
( 3 -4 1 )
Cov(X) = ( -4 10 -2 )
( 1 -2 3 )
Let Y = X1 + 2X2 + 3X3 and Z =
X2 +4X3 . Recall that Var(Z) means the
variance of Z.
(i) Find Var(X1) and Var(X2).
(ii) Find Var(Y) and Var(Z) .
(iii) Find the covariance Cov(Y,Z) = E(YZ) .
(iv) Let W be the random vector (X_1 X_3)'. Find Cov(W).
5. Consider the blood glucose measurements for three visits by each
of 50 female subjects in Table 3.8 on page 80 in the text. (See
BloodGlucose.dat on the Math 439 Web site.)
(i) Write a SAS program that uses SAS's matrix language (Proc IML) to
compute the sample mean vector and sample covariance matrix of the six
variables in the table. Partition the mean vector and sample covariance
matrix into two groups with three variables each, as in Table 3.8 and
Section 3.8 in the text.
Did you get the same answer as in the back of the book? (This problem
is similar to problem 3.22 in the text, whose answer is in the back
of the book.)
(ii) Have your SAS program also calculate the sample correlation
matrix of the six variables.
(iii) In the same program, also use SAS's Proc Corr to compute the
sample mean vector, sample covariance matrix, and sample correlation
matrix. Do you get the same answers?
(Hint: See Calcium.sas
or
Calcium2.sas
on the Math439 Web site.)
(iv) Which set of three variables (either Before=y1,y2,y3 or
After=x1,x2,x3) appear to be more highly correlated among themselves for
the 50 subjects? Does this suggest that fasting blood-glucose levels are
more uniform among these subjects, or that blood-glucose levels after
sugar intake are more uniform?
6. Do problem 4.16 (page 108) with X, mu, and Sigma given by
( 1 ) ( -2 ) ( 5 3 | -1 -1 )
xx = ( 2 ) mu = ( 3 ) Sigma = ( 3 7 | 0 -1 )
(----) ( ----------------- )
( -1 ) ( -1 0 | 2 3 )
( 5 ) ( -1 -1 | 3 5 )
Either do this either by hand, or else include the calculations in your
SAS program for Problem 5. (Warning: In the book's notation, the
upper-left corner of Sigma is S_yy and the upper-right corner is S_yx,
which is the transpose of S_xy. Note that the answer to problem 4.16 is
also given in the back of the book.)
Top of this page