MA 780: Midterm Test and Solutions

1. Given (x0 , y0 ), (x1 , y1 ), , (xn , yn ). Write down the following:

(a) the formula of Lagrange polynomial interpolation.
(b) the Newton polynomial interpolation using the divided difference.
(c) the error estimate.
(d) Show that

n
X

li (x) = 1 if n > 0, where li (x) is the Lagrange polynomial.

i=0

Let the data be (1, 0), (1, 3), (0, 1), write down one of the two polynomial interpolations.
Solution: (a) Lagrange polynomial interpolation:
pn (x) =

n
X

li (x)yi ,

li (x) =

i=0

n
Y

x xj
x
xj
j=0,j6=i i

(b) Newton polynomial interpolation:

pn (x) =

n
X

f [x0 , x1 , , xi ] i (x),

i (x) = (x x0 )(x x1 ) (x xi1 )

i=0

(c) We know that f (x) pn (x) =

f (n+1) ()
n+1 (x).
(n + 1)!

(d) Take f (x) 1, then f (n+1) () 0 when n > 0. That is, the interpolation is exact.
Therefore, we use the Lagrange polynomial interpolation to get
1 f (x) = pn (x) =

n
X

li (x).

i=0

Using Newton divided difference, we get

5
3
p2 (x) = (x 1) + (x 1)(x + 1).
2
2
2. Find the divided differences:
(a) f [0, 1, 2, , 7] if f (x) = x7 + x3 + 1.
(b) f [0, 1, 2, , 7, 8] if f (x) = x7 + x3 + 1.
(c) f [x0 , x1 , x2 , , xi1 , xi+1 , , xn1 , xn ] if f (x) = li (x).
Solution: (a) Using f [x0 , xi , , xi ] =
(a)
(b)

f (i)
i! ,

we get

7!
=1
7!
f (8)
= 0.
f [0, 1, 2, , 7, 8] =
8!
f [0, 1, 2, , 7] =

(c) Using f [x0 , x1 , , xi ] =

i
X

f (xj )

(x )
j=0 i+1 j

= 0 since li (xj ) = 0 if i 6= j.

3. (a) Give the definition of a cubic spline S3 (x) given the nodal points x0 , x1 , , xn . What
are the three common boundary conditions?
(b) What is a B-spline interpolation? List some properties of a B-spline interpolation.
Solution: (a) A cubic spline S3 (x) is piecewise cubic function between [a, b], S3 (x)
has up to second order continuous derivatives. The three commonly used boundary
conditions are: (1) Natural spline: S3 (a) = S3 (b) = 0. (2) Clamped spline: S3 (a) =
(k)
(k)
f (a) and S3 (b) = f (b) are given. (3): Periodic spline: S3 (a) = S3 (b) for k = 0, 1, 2.
(b) A B-spline interpolation has the following form
S3 (x) =

n1
X

i Ni (x),

S3 (xi ) = f (xi )

i=3

where Ni (x) = g[xi , xi+1 , xi+2 , xi+3 , xi+4 ; t] with g(x, t) = (x t)3+ . 0 Ni (x) 1 and
has local support.
(x)|2 dx b |f (x)|2 dx
Also if S3,f is the cubic spline interpolation, then we have ab |S3,f
a
R b
R b
2

2
and a |f S3,f (x)| dx a |f (x) S3 (x)| dx for any spline function S3 (x).

4. Write down the Newton-Cotes quadrature closed, and open formulas and give an error
estimate.
Solution:
b

f (x)dx h

n
X

wi f (xi )

i=0

wi =

n+1

wi =

tk
dt
ik
k=0,k6=i
Y

n
Y

for closed formula with h =

tk
dt
ik
k=0,k6=i

ba
n

for open formula with h =

ba
n+2

The error estimate is

En (f ) =

f (x)dx In (f ) =

Mn
hn+3 f (n+2) () if n is even.
(n + 2)!
Kn
hn+2 f (n+1) () if n is odd.
(n + 1)!

5. (a) Write down (a) mid-point rule; (b) trapezoidal rule; (c) Simpson rule to evaluate the
integral

xi +h

xi

x3 dx and find the error estimates in terms of h.

(b) Find out and show the algebraic precision of the trapezoidal rule.
Solution:
Z

xi +h

xi
Z xi +h

x3 dx h (xi + h/2)3


h 3
xi + (xi + h)3
2
xi
Z xi +h


h
x3 dx
x3i + 4(xi + h/2)3 + (xi + h)3
6
xi

x3 dx

3
The error estimates for he mid-point rule and the trapezoidal rule are both O(h
 ) and for
Rb k
ba
k
k
the Simpson rule, the error is 0. It is easy to check that a x dx = 2 a + b for k = 0,

and k = 1, but ab x2 6= ba
a2 + b2 , therefore the algebraic precision of the trapezoidal
2
rule is 1. The algebraic precision of the mid-point rule and the Simpson rule are 1 and 3,
respectively.
R

6. Set-up a system of equations for the coefficients 1 , 2 , x1 and x2 so that the following
quadrature
Z

f (x)dx 1 f (x1 ) + 2 f (x2 )

has as high algebraic precision as possible. What is likely the algebraic precision? Note:
You do not need to solve the system of equation.
Solution: Since there are 4 parameters, we can set 4 equations
x0 :
x1 :
x2 :
x3 :

1
Z 1

1
Z 1
1
1

dx = 2 = 1 + 2
xdx = 0 = 1 x1 + 2 x2
x2 dx =

2
= 1 x21 + 2 x22
3

x3 dx = 0 = 1 x31 + 2 x32

The algebraic precision of the quadrature formula is likely to be 3.

7. If A(h) = A0 + Ch3 + O(h5 ), find the formula of the Richardson extrapolation.
Solution:
Bh =

Bh =

A( h) 3 A(h)
1 3

8A(h/2) A(h)
A(h/2) A(h)
=
,
1 3
7
1 (2)

1
.
2