Lagrange Method of

Interpolation
 What is Interpolation ?

Given (x0,y0), (x1,y1), …… (xn,yn), find the

value of ‘y’ at a value of ‘x’ that is not given.

2 http://numericalmethods.eng.usf.edu
 Interpolants
Polynomials are the most common
choice of interpolants because they
are easy to:

Evaluate
Differentiate, and
Integrate.

3 http://numericalmethods.eng.usf.edu
 Lagrangian Interpolation

Lagrangian interpolating polynomial is given by

n
f n ( x)   Li ( x) f ( xi )
i 0

where ‘ n ’ in f n (x) stands for the n th order polynomial that approximates the function y  f (x)

given at (n  1) data points as x0 , y 0 , x1 , y1 ,......,  x n 1 , y n 1 , x n , y n  , and

n x  xj
Li ( x)  
j 0 xi  x j
j i

Li (x) is a weighting function that includes a product of (n  1) terms with terms of j  i

omitted.

4 http://numericalmethods.eng.usf.edu
 Example-1
Find the root, correct to three decimal places and lying
between 0 and 0.5, of the equation
4e x sin  x   1  0
Solution: Let
f  x   4e x sin  x   1  0
We have f  0  1 and f  0.5  0.163145
Therefore,
x1  0.25

Since f  0.25  0.22929 , it follows that the root lies between

0.25 and 0.5.
Therefore,
0.75
x2   0.375
2
 The successive approximation are given by

x3  0.3125 x4  0.3438 x5  0.3594

x6  0.3672 x7  0.3711 x8  0.3692

x9  0.3702 x10  0.3706 x11  0.3704

x12  0.3705

Hence the required root is 0.371, correct to three decimal

places.
 Example-2
Certain corresponding values of x and log10 x are

xi f(xi)
Form formula
300
300, 2.4771 ,
2.4771

304 2.4829

305 2.4843
307 2.4871

Find the value of log10 301

Ln  x  
 x  x1  x  x2 ....... x  xn 
y0 
 x  x0  x  x2 ....... x  xn 
y1  ........
 x0  x1  x0  x2 ........ x0  xn   x1  x0  x1  x2 ........ x1  xn 
  301  304  301  305 301  307   301  300  301  305 301  307 
log10 301   2.4771   2.4829
300  304300  305 300  307   304  300 304  305 304  307
 301  300  301  304  301  307   301  300  301  304  301  305
  2.4843   2.4871
305  300305  304305  307  307  300 307  304 307  305

 3 4  6  1 4  6 
log10 301   2.4771   2.4829 
 4  5 7   4  1 3
1 3 6  1 3 4 
  2.4843   2.4871
 51 2   7  3 2 

 1.2739  4.9658  4.4717  0.7106

 2.4786

8
 Find the Lagrange interpolating polynomial of degree 2
approximating the function y  ln x defined by the following
table of values.
2 0.69315
2.5 0.91629
3.0 1.09861

Hence determine the value of ln 2.7 .

We have
l0  x  
 x  2.5 x  3.0 
 2 x 2  11x  15
 0.5 1.0 
Similarly, we find
l1  x     4 x 2  20 x  24 
and
l2  x   2 x 2  9 x  10
Hence
L2  x    2 x 2  11x  15  0.69315   4 x 2  20 x  24   0.91629 
  2 x 2  9 x  10  1.09861

 0.08164 x2  0.81366 x  0.60761

Which is required quadratic polynomial.

Putting x  2.7, in the above polynomial, we obtain

ln 2.7  L2  2.7   0.08164  2.7   0.81366  2.7   0.60761  0.9941164

2

Actual value of ln 2.7  0.9932518, so that

Error  0.0008646
 Program for Lagrange’s Method
Dimension p(50),x(50),f(50)
open(unit=5,file='lagr.in',status='unknown')
open(unit=6,file='lagr.res',status='unknown')
read(5,100)n,xx
Write(6,100)n,xx
read(5,101)(x(i),f(i),i=1,n)
write(6,101)(x(i),f(i),i=1,n)
100 format(I5, f10.4)
101 format(2f10.4)
Do 10 j=1,n
P(j)=1.0
Do 11 i=1,n
If(i-j)5,11,5
5 P(j)=p(j)*((xx-x(i))/(x(j)-x(i)))
11 continue
10 continue
Sum=0.0
Do 15 k=1,n
Sum=sum+p(k)*f(k)
15 continue
Write(6,102)xx,sum
102 format(2e13.4)
close(unit=6)
close(unit=5)
Stop
end
Hermite Method of
Interpolation
 Hermite’s Interpolation Formula

 In Hermite interpolation formula both the function and its first derivative values
are to be assigned at each point of interpolation. This is referred to as Hermite’s
interpolation formula.
 Given the set of data points  xi , yi , yi'  , i  0,1,.....n it is required to
determine a polynomial of the least degree, say H 2 n 1  x 
Such that
H 2 n 1  xi   yi
H 2' n1  xi   yi' (1)
i  0,1, 2,......n

We have here  2n  2  conditions and therefore the number of coefficients to

determined is  2n  2  and the degree of the polynomial is  2n  2 
 In analogy with the Lagrange interpolation formula, we seek a representation of
the form
n n
H 2 n 1  x    ui  x  yi  vi  x  yi'
i 0 i 0

Where ui  x  and vi  x  are polynomials in x of degree  2n  2  . Using equation

(1), we obtain
1  i  j
ui  x j    ; vi  x   0,
 0  i  j
ui'  x   0 for all i
1  i  j
vi'  x j   
0  i  j

Where ui  x  and vi  x  are polynomials in x of degree  2n  2  , we write

  x  x 
li  x  
2
ui  Ai 
 x  x 
li  x  
2
vi  Bi 

 x  ai x  bi  
li  x  
2
ui  
 x  ci x  d i  
li  x  
2
vi  

Using above conditions we obtain

 ai xi  bi   1
 ci xi  d i   0
ai  2li'  xi   0
ci  1
From the last equation we get
ai  2li'  xi 
bi  1  2 xi li'  xi 
ci  1 and di   xi
On substituting we get

ui  x   1  2  x  xi  l  xi  li  x  

' 2
i
(2)

vi  x    x  xi  
li  x  
2


Then Hermite interpolation formula becomes

n n
H 2 n1  x    1  2  x  xi  l  xi   li  x   yi   x  xi  li  x  yi'
' 2 2
i
i 0 i 0

(3)
 Example-1
Example 1. Determine the Hermite polynomial of degree 5, which fits the following data
and hence find an approximate value of y  ln 2.7 .
x y  ln x y '  1/ x

2.0 0.69315 0.5

2.5 0.91629 0.4000
3.0 1.09861 0.33333
The polynomials li  x  have already been computed in Example 3. These are

l0  x   2 x2  11x  15, l1  x     4 x 2  20 x  24  , l2  x   2 x 2  9 x  10

We therefore obtain

l0'  x   4 x  11, l1'  x   8x  20, l2'  x   4 x  9

Hence
l0'  x0   3, l1'  x1   0, l2'  x2   3
Equations 2 give
v0  x    x  2   2 x 2  11x  15 ,
2
u0  x    6 x  11  2 x  11x  15 ,
2 2

v1  x    x  2.5  4 x 2  20 x  24  ,
2
u1  x    4 x  20 x  24  ,
2 2

v2  x    x  3  2 x 2  9 x  10  ,
2
u2  x   19  6 x   2 x  9 x  10  ,
2 2

Substituting these expressions in equation 3, we obtain the required Hermite polynomial

H 5  x    6 x  11  2 x 2  11x  15   0.069315 

2

  4 x 2  20 x  24   0.91629 

 19  6 x   2 x 2  9 x  10  1.09861
2

  x  2   2 x 2  11x  15   0.5 
2

  x  2.5   4 x 2  20 x  24   0.4 
2

  x  3  2 x 2  9 x  10   0.33333
Putting x  2.7 and simplifying, we obtain

ln  2.7   H 5  2.7   0.993252

Which is correct to six decimal places. This is therefore a more accurate result than that
obtained by using the Lagrange interpolation formula.
 Program for Hermite Method

 Dimension x(50),f(50),fd(50),al(50),alp(50)
 open(unit=5,file='lagr.in',status='unknown')
 open(unit=6,file='lagr.res',status='unknown')
 read(5,*)n,xx
 Write(6,*)n,xx
 read(5,*)(x(i),f(i),fd(i),i=1,n)
 write(6,*)(x(i),f(i),fd(i),i=1,n)
 call poly(x,f,n,xx,al,alp)
 term=0.0
 do 1 i=1,n
 term1=(1-2*(xx-x(i))*alp(i))*al(i)*al(i)*f(i)
 term2=(xx-x(i))*al(i)*al(i)*fd(i)
 1 term=term+term1+term2
 write(6,*) xx,term
 close (unit=6)
 close (unit=5)
 Stop
 end


20
 Program continued
 Subroutine poly(x,f,n,xx,p,pd)
 Dimension p(50),x(50),f(50),pd(50),pp(50)
 Do 10 j=1,n
 P(j)=1.0
 Do 11 i=1,n
 If(i.eq.j)go to 11
 p(j)=p(j)*(xx-x(i))/(x(j)-x(i))
 11 continue
 10 continue
 Do 20 j=1,n
 xxx= x(j)
 Pp(j)=1.0
 pd(j)=1.0
 Do 112 i=1,n
 If(i.eq.j)go to 15
 pp(j)=pp(j)*(xxx-x(i))/(x(j)-x(i))
 15 continue
 ss=0.0
 Do 13 i=1,n
 If(i.eq.j)go to 13
 ss=ss+1.0/(xxx-x(i))
 13 continue
 pd(j)=pp(j)*ss
 20 continue
 return
 end

21
 Program continued
 Subroutine poly(x,f,n,xx,p,pd)
 Dimension p(50),x(50),f(50),pd(50),pp(50)
 Do 10 j=1,n
 P(j)=1.0
 Do 11 i=1,n
 If(i.eq.j)go to 11
 p(j)=p(j)*(xx-x(i))/(x(j)-x(i))
 11 continue
 10 continue
 Do 20 j=1,n
 xxx= x(j)
 Pp(j)=1.0
 pd(j)=1.0
 Do 15 i=1,n
 If(i.eq.j)go to 15
 pp(j)=pp(j)*(xxx-x(i))/(x(j)-x(i))
 15 continue
 ss=0.0
 Do 13 i=1,n
 If(i.eq.j)go to 13
 ss=ss+1.0/(xxx-x(i))
 13 continue
 pd(j)=pp(j)*ss
 20 continue
 return
 end

22
Least Square Method
 Introduction
 In experimental work, we often encounter the problem of fitting
a curve to data which are subject to errors.

 The strategy for such cases is to derive an approximating

function that broadly fits the data without necessarily passing
through the given points.

 The curve drawn is such that the discrepancy between the data
points and the curve is least.

 In this least squares method, the sum of the squares of the errors
is minimized.

24
 The procedure for least square curve fitting
 Let the set of data points be  xi , yi  , i  1, 2,3..., m .
 Let the curve given by Y  f  x  be fitted to this data.
 At x  xi , the given ordinate is yi and the corresponding
value of the fitting curve is f  xi  .
 If ei is the error of approximation at x  xi , then we have

ei  yi  f  xi 
If we write
S   y1  f  x1    y2  f  x2   ....   ym  f  xm 
2 2 2

S  e12  e22  ......  em2 ,

Then the method of least squares consists in minimizing S i.e., the
sum of squares of errors.

25
 Fitting a straight line
Let Y  a0  a1 x be the straight line to be fitted to the given data,
viz.  xi , yi  , i  1, 2,3..., m. Then, the corresponding equation, we
have,
S   y1   a0  a1 x1    y2   a0  a1 x2   ....   ym   a0  a1 xm 
2 2 2

For S to be minimum, we have

S
 0  2  y1   a0  a1 x1    2  y2   a0  a1 x2   ....  2  ym   a0  a1 xm 
a0
And
S
 0  2 x1  y1   a0  a1 x1    2 x2  y2   a0  a1 x2   ....  2 xm  ym   a0  a1 xm 
a1

The above equations can be simplified to

ma0  a1  x1  x2  ...  xm   y1  y2  ....  ym

26
 And
a0  x1  x2  ...  xm   a1  x12  x22  ...  xm2   x1 y1  x2 y2  ....  xm ym

The above equations can be written as

m m
ma0  a1  xi  yi
i 1 i 1
And
m m m
a0  xi  a1  x  xi yi 2
i
i 1 i 1 i 1

On solving these we get

m m m
m xi yi   xi . yi
a1  i 1 i 1 i 1
2
 m
m

m x    xi 
2
i
i 1  i 1 

And then
a0  y  a1 x
27
 2S 2S
Since
a02
and a12
are both positive at the points a0 and a1 , it
follows that these values provide a minimum of S.

The last equation can be written as

y  a0  a1 x
Which shows that the fitted straight line passes through the centroid
of the data points.
The correlation coefficient (cc) can be defined as:

Sy  S
cc 
Sy

Where
m
S y    yi  y 
2

i 1

If cc is close to 1, then the fit is considered to be good.

28
 Example-1
Find the best value of a0 and a1 if the straight line Y  a0  a1 x
is fitted to the data  xi , yi  :

X 1 2 3 4 5
Y 0.6 2.4 3.5 4.8 5.7

and also find the correlation coefficient.

 yi  y   yi  a0  a1 xi 
2 2
xi yi xi2 xi yi

1 0.6 1 0.6 7.84 0.0754

2 2.4 4 4.8 1.00 0.0676
3 3.5 9 10.5 0.01 0.0100
4 4.8 16 19.2 1.96 0.0196
5 5.7 25 28.5 5.29 0.0484
15 17.0 55 63.6 16.10 0.2240
29
 From the table, we find x  3 and y  3.4 , then
5  63.5  15 17 
a1   1.26
5  55  225
Therefore,
a0  y  a1 x  0.38

16.10  0.2240
The correlation coefficient=  0.9930
16.10

30
 Linearization of Nonlinear Laws
 The data given may not always lie in a linear relationship form.
This can be ascertained from a plot of the given data.
 If a nonlinear model is to be fitted, it can be conveniently
transformed to a linear relationship. This can be done by:
Case-I: xy a  b
Taking log of both sides, we get

log10 x  a log10 y  log10 b

In this case ,we put

log10 x  X log10 y  Y
And
1 1
log10 b  A0 and   A1
a a
31
 Then we get

Y  A0  A1 X
Case-II:
y  aebx
The above equation can be written as

ln y  ln a  bx
Y  A0  A1 X
where
Y  ln y A1  b
and
A0  ln a X x

32
 Example-2
Using the method of least squares, find constants a and b such that
the function y  aebx fits the following data:

X 1.0 3.0 5.0 7.0 9.0

Y 2.473 6.722 18.274 49.673 135.026

We have,
y  aebx
Therefore
ln y  ln a  bx
Y  A0  A1 X
Where
Y  ln y A1  b
and
A0  ln a X x

33
 The table of values is given below

X Y=ln y X2 XY
1 0.905 1 0.905
3 1.905 9 5.715
5 2.905 25 14.525
7 3.905 49 27.335
9 4.905 81 44.145
25 14.525 165 92.625

We obtain
X 5 and Y  2.905

5  92.625  25 14.525 
A1   0.5  b
5 165  625
34
 Then
A0  Y  A1 X  2.905  0.5  5  0.405

Hence,
a  e A0  e0.405  1.499
It follows that the required curve I of the form

y  1.499e0.5 x

35
 Curve Fitting by Polynomials
Let the polynomial of the nth degree,

Y  a0  a1 x  a2 x2  ......  an xn
Be fitted to the data points . We then have
S   y1   a0  a1 x1  a2 x12  .....  an x1n     y2   a0  a1 x2  a2 x22  .....  an x2n  
2 2

....   ym   a0  a1 xm  a2 xm2  .....  an xmn  

2

Equating to zero the first partial derivatives and simplifying, we obtain the normal
derivatives:
ma0  a1  xi  a2  xi2  .....  an  xin   yi
a0  xi  a1  xi2  .....  an  xin 1   xi yi
. . . .
. . . .
a0  xin  a1  xin1  .....  an  xi2 n   xin yi

where the summations are performed from i=1 to i=m.

36
 The Previous equation constitutes (n+1) equations in (n+1) unknowns, and hence
can be resolved by

Example-3
Fit a polynomial of the second degree to the data points (x,y) given by.

X 0 1 2
Y 1 6 17
For n=2, the table is as follows:
X Y X2 X3 X4 XY X2Y

0 1 0. 0 0 0 0

1 6 1 1 1 6 6

2 17 4 8 16 34 68

3 24 5 9 17 40 74

37
 The normal equations are

3a0  3a1  5a2  24

3a0  5a1  9a2  40
5a0  9a1  17 a2  74
On solving the above equations, we obtain

a0  1, a1  2 and a2  3
The required polynomial is given by

Y  1  2 x  3x 2
and it can be seen that this fitting is exact.

38
 Multiple linear least squares
Suppose that z is a linear function of two variable x and y. If the
function z  a0  a1 x  a2 y is fitted to the data
 z1 , x1 , y1  ,  z2 , x2 , y2  .......  zm , xm , ym 
Then the sum
m
S    zi  a0  a1 xi  a2 yi 
2

i 1

should be minimum, for this, we have

S
 2  zi  a0  a1 xi  a2 yi   0
a0

S
 2 xi   zi  a0  a1 xi  a2 yi   0
a1

S
 2 yi   zi  a0  a1 xi  a2 yi   0
a2
39
 On simplifying, we get

ma0  a1  xi  a2  yi   zi

a0  xi  a1  xi2  a2  xi yi   zi xi

a0  yi  a1  yi xi  a2  yi2   zi yi

From which a0 , a1 and a2 can be determined.

40
 Example-4
Find the value of a0 , a1 and a2 , so that the function
z  a0  a1 x  a2 y
is fitted to the data  x, y, z  given below.
(0,0,2), (1,1,4), (2,3,3), (4,2,16) and (6,8,8)

We form the following table of values

X Y Z X2 XY ZX Y2 YZ
0 0 2 0 0 0 0 0
1 1 4 1 1 4 1 4
2 3 3 4 6 6 9 9
4 2 16 16 8 64 4 32
6 8 8 36 48 48 64 64
13 14 33 57 63 122 78 109
41
 The normal equations are

5a0  13a1  14a2  33

13a0  57a1  63a2  122
14a0  63a1  78a2  109

On solving the above equations, we get

a0  2 a1  5 a2  3

42