2 Finite Difference Operators and

Interpolation with Equal Intervals

2.1. Finite Differences

Introduction. Let y =fx) be a function of x and corresponding to every value of xin the do
main a SxS6, there exist one or more values of y. In case f(x) is specified by a given explicit formula,
we can compute the value or values of fx) corresponding to some given value of x by simply sub
stituting the value of x in the fornula. However, if the function f(x) is not known specifically, even
then the values of f(x) can be found with the help of calculus of finite differences provided some
values
values of fx) are known. Thus the calculus of finite differences deals with the changes in the
of thefunction (dependent variable) due tochanges in the independent variable.
function assumes,
Again through this, we study the relations of the values which theunequal.
or
whenever the independent variable changes by finite jumps whether equalintervals. In this chap
ter, we shall study the case when independent variable changes by equal
2.2. Some Definitions
of x, then the independent variable x is
1. Argument and Entry. Let y =fx) be a functionfunction is called entry.
called the argument of the function and the value of the
let the corresponding values of y at
For example, ify=f(x)° be the function of x and
a, a +h, a +2h, ...are called the arguments and
A= a, a +h, a + 2h...be y, Y,+h, Ya +2h.. thenentries.
corresponding
Yu Ya +h» Ya + 2h:..are called the
between consecutive values of the argument ' is
2. Interval of Differencing. The differenceexample, interval of differencing is ).
Lalled the intervalof differencing. (In the above
Forward Differences. Let y= f(x) be the function of x. Then forward difference of fx) is
3.
denoted by Af(x) and is defined as ...(1)
Afx)=fx +h) - fr).
called forward difference operator.
where h is the interval of differencing.. Ais
Backward Difference. The bachward difference of flx) is denoted by Vfx) and is defined as
4. ..2)
Vfr) =f*)-fu-h),
operator.
Vis called backward difference defined as
Operator E. The operator E, called the displacement or shift operator is
5, Shift ...(3)
Efr)=fr h).
2.3. Finite Differences corresponding
+ nË be the entries of the function y, =fx) [or y = fx))
Let y, ya +b,Ya +2h, ... ya differences
arguments a,a + h, a+ 2h.,....,.a+ nh respectively, then the
to the in- 1
-Yas ya+ 2h- Ya+ h» Ya +3h Ya + 2/h**, ya +nh - Ya+
Ya +h
differences ofy=fa) and we denote these differences by
are called the first forward
Ay, Aya + h AYa +2h» , AYa +(n - 1y
fr) is denoted by y,, thus fla)=y, fa + h)=y, .h, fla + 2h) =y,+2, etc.
For the sake of convenience,
26 NUMERICAL ANALYSIS

Ay, =Ya+h -Ya

Aya +h=Ya +2h-Ya +h
L.e., Aya +2h =Ya +3h - Ya+ 2h ....1)

Aa +n - 1)h =Ya+nh -Ya+(n - 1)h

The differences of the first forward differences given by (1) are called second forward ditfer.
ences and are denoted by Ay,Ay h..etc. Therefore we have
Ay,= Aya +h-Aya
Aya +h= Aya +2h -Aya +h
Proceeding in this way, we get differences of higher order viz.
Ay, 'ya +hy...
Ay, Aya +hys.
and s0 on.
In general the nth forward difference is given by

Note : 1. ya, the first entry is termed the leading term and the differences Ay A"ya ..."y,
are known as the leading differences.
2. It should be kept in mind that Ais an operator and A? is not the square of the operator A, but
it is repetition of the operator Aete.
Atable showing the diagonal forward differences is called the diagonal forward difference
table. Let yo, y1, '2, Y3,Y be the entries corresponding tothe arguments 0, 1, 2, 3, 4 of a function
fx). Then diagonal forward difference table may be obtained as :
Table 1.Diagonal Forward Diference Table
Argument Entry First Second Third Fourth
Diferences Ay Differences A'y Differences A'y Differences Ay

Y1-Yo = Ayo
1 AyË - Ayo =A'yo
'y- 'yo =Ay
2 Ay, - Ay =A'y,

3 Ayg - Ay, =A'y,

4 Y4

Here y, is the leading term and Ao 4yo, A'yo, A'yo are the leading differences.
 DIFFERENCE
FINITE OPERATORS AND INTEROLATION 27

Remark:
Various steps employed in constructing table 1.
1. Column() contains the values of the
(andthe table given in example
9Column (i) contains the independent variablex.
8. Column () 1s caleulatedcorresponding
value of v for given value of x.
by subtracting each value fr) in
value off). column (ii) from the preceering
4. The operation of step 3is applied on the
ll allthe figures in a column become constant. ffgures of column Gi)) toget column (iv) and so on
SOLVED EXAMPLES
Example 1. Construct aforuward difference table for the
following data :
45 50 55 60 65
fx): 2871 2-404 2-083 1-862 1-712
Solution.
The forward difference table is
y=fr) Ay Ay A'y
45 2-871.

-0-467!
50 2.404 0-146
0-321 0-046
55 2-083 0-100 0-017
-0-221 -0-029
60 1-862 0-071
-0-150
65 1-712

Here leading term y45 = 2-871 and leading differences are

Ay4s =-0:467, N'y4s= 0146, A'yg=- 0-046, Ay; =0017.
2.4. Backward Differences
Backward difference operator Vis defined as
Vfx) =fx)-fr-h).
Thus if yu- h» YasYa +hy are the entries corresponding to the arguments a-h, a, a +h,..
respectively for the function y =fu), then first backward differences are
Vy,=ya -Ya-h
Vya +h =Ya +h-Yu

Vya +nh =Ya + nh -Ya +(n - 1)h

 28 NUMERICAL ANALYSIS

Second backward differences are V'y,h = Vy+h - Vya

Vya+ 2, = Vya+2h

Vy,+nh = Vya+nh - Vya+ (n -1)%

In general, nth order backward difference is given by
V"ya+nh = " y , nh - 7ya+ (n - 1h
Let Yo» y Y2s V3, , be the functional values of the function f(x) for x = 0, 1, 2, 3, 4. The back
ward difference table is given below :
Diagonal Backward Difference Table
Vy Vy Vy vy
Yo

Y1-Yo = Vy

J2-Y) = Vy
2

Ys-y2 = Vy3

Y4-Y3= Vy4
4 y4

In backward difference table, the differences in each column start from bottom of the table and
Vy4, Vy4, V'y, Vy,become the leading backward differences.
Example 2. Construct horizontal backward difference table for y= sinx when
x=15, 20°, 25°, 30° and find the value of sin 10° from the table. It is being assumed that third
differences are constant.
Solution.

Vy v'y
10
0-2588 -0-0852 0-0832- (-0-0020) 0-0026 -(-0-0006)
=0-1736 =0-0852 - 0-0006
=-0-0020.
15 0-2588
0-0832.
20 0-3420 -0-0026
0-0806 - 0-0006
25 0-4226 0-0032
0-0774
30 0-5000
FINITE DIFFERENCE
OPERAORS AND INTEROLATION
vy being constant,

v'y0=- 0-0006
V'y=y6-' yo
=-0-0026 -(-00006) =0-0020

= 0-0832
-(-0-0020) = 0-0852
Y1o =Y15 - Vy0= 0-2588- 0-0852 = 0-1736
sin 10 =0.1736
2.5. Relation of Operators E, Aand
2.5.1. Prove that E 1+ A.
Proof. We know that
Ef(¢) =fx +h) ...(1)
and
Af(o) =f( +h) -f()
Af(*) =Ef)-f() (Using (1)I
A=E-1
or
E=1+A.
Note. (i) While usingE =1+ ,it should be kept in mind that 1(unity) usedhere is an
such that 1.f() =f(r) identity operator
(iü) E-f() =f(r- h)
Ef() =fu- 2h)

E-"f() =fu- nh)

2.5.2. Prove that V=l-E-1
Proof. We know that Vf() =f(e)-f(x - h) =f(o) -E-f(u)
Vf(*) =(1-E-}f(0)
V=1-E-1

2.5.3. Prove that V=AE1

Proof. We know that Vf(r) =f()-f(x- h)
= Af(- h)
=AE'f()
V= AE-1
Z.6. Fundamental Theorem of Difference Calculus (Statement)
ff() is a polynomial of degree n in x, then the nth difference of f() are constant
and (n + 1)th difference are zero. (Proof of the theorem is beyond the scope of the book)
30 NUMERICAL ANALYSIS

2.7. Interpolation With Equal Intervals

2.7.1. Interpolation
It is the process of estimating the value of the function (or entry) for any intermediate value of
the variable (or argument) with the help of its given set of values (which may be at regular intervals
or at unequal intervals). Let us assume that the function ý=f) is known for certain values ofx say
for o. *}, *2,.., as fro), fu), fm)...,). With the help of these given values of the function, the
process of finding the value of f) for some other value of x, lying between the values
o, 1, 2 . , is called interpolation.
The exact form of the function f(w) is generallyunknown or even if known it is usually of a
complicated nature. Hence for interpolation purpose,we have to depend on finite difference
methods. All finite difference methods are based on the assumption that :
() The given function is a polynomial in &. The degree of the polynomial depends upon the
number of given values of the function, In general, ifn +1values are given, we assume that () is
a polynomial of degree n unless stated otherwise.
(z)The values of the function should be either in increasing or in decreasing order i.e., there
should not be sudden jumps or falls in the values of thefunction in the interval under consideration.
In this section we shall discuss the Newton's forward and Newton's backward formulae for
interpolation when the values of arguments x are equally spaced.
2.8. Newton-Gregory Formula for Forward Interpolation [K.U.2001, 96; M.D. U. 1998]
Let y = fia)' be a function of x which assumes the values fa), fa +h), fa + 2h),.., fa +nh) for
n+lequidistant välues a, a + h,a +2h ..a+nh ofx. Let us assume that fr) is a polynomial inzof
degree n and may be written in the form

fa) = Ap t A,(*- a) +Az(* - a)x -a+h) + Ag(r- a)x -a+ h)(x -a+ 2h) +...
+ A,(x-a)(x - a+ h)...(« a + (n-1)) ..1)
where A, Aj....A, are constants to be determined.
Puttingx = a, a +h, a +2h,.. a +nh in (1), we get
Fof x 0,. fa) = A
For x=a+ h, fla + h) =Ap t+ A,.h
Ajh =fa +b) - Ay=fa +h) -fa) =Afa)
A, a)
For x= a+ 2h, f(a +2h) =Ap +A,.2h +A(2h).h
A. 2h =f(a +2h) - Ap- A,.2h
-fla+2h) - fa) - Afia)
h
2h

=fla +2h) -fla) - 2fla + h)-fa)) =fla + 2h)- 2f(a + h) +fla)

=fa + 2h) -fa + h) - fa + h) -fla))
= Afa + h) - sfa) = Afla).
FINITE DIFFERENCE
OPERATORS AND INTERPOLATION 31

A, =
(2)! h2
Similarly, we get Ag =
(3) ! h3

A, =
(n)! h"
Putting the values of Ao, A1, Ag, . A, in (1), we
get
flx) =fla) +(* - a) + (*- a)(x-a+h)
(2) ! 12
+( - a)(x - a+ h)..(x -- a + (n - 1) h) Nf ) ...(2)
(n) !h
which is the Newton-Gregory Formula for Forward
mainly for interpolating the values of fx) near the beginningInterpolation. This formula is used
(x>a) of a set of tabulated values.
Put u= i.e., x= a+hu
h
(where x is the value of the variable at which the value of function is to be
interpolated)
I-a = uh
Also x-a+h =(u- 1)h,
x-a+2h = (u- 2)h,

x-a+(n - 1) h =(u -n- 1) h,

Now putting all these values in (2), we have
fla + hu) =fa) + (uh) +(uh)(u -1).
h (2) !. h?
t... +(uh)(u - 1)h....(u - n-1)h. A'fa)
(n)!. ha
=fa) +usfa) +uu-1)
2!
2 Aol uu= 1)(u 2)
3! fla)
su - 1)(w-
+... +
2)...s-n-Dfa)
n!
The above result can be written as
u) u?) s)
fa + uh) =fa) + 114(0) t a)+.... n!
where u=u.(u - 1).(u - 2)..(u -r+ 1) is factorial polynomial of order r.
This is the form in which Newton-Gregory formula for forward interpolation is usually
written.
 NUMERICAL ANALYSIS
32

Interpolation
Newton Gregory Formula for Backward
The backward difference for any function fix) is defined as
Vf(*) = f(x) -f(r - h),
h being the interval of differencing.
which assumes the values fa + nh), fa +n- 1h),.... f(a) for
Let y =f)be a function ofx
given n +1values of the function.
*=a+nh, a +n-1h,.,a +h. a ie.. we are
polvnomial of nth degree. Let the polynomial be
.1nen to n + lobservations, we can ft a
T() = Ao t A,(x -a+nh) + A(x -a+nh)(* - a + (n - 1)h t ..*
l)h....( -a+ h) ...1)
+ A,,(a -a+ nh)(x -a + (n-
where A, A,, A, .A, are constants to be determined.
Putting x=a+nh in (1), we get fa + nh) = Ap
Puttingx= a+n -1h, we get
fla +n -1h) = A + Aj( - h)
+n -1h)_ Vfla + nh)
A,=fa +nh)-fa

Puttingx=a+n - 2 h, we get
fa +n-2 h) = Ap + Aj(-2h)+ A(-2h)( -h)
Or 2h" A, =fla +n- 2 h) - Ap +2h.A,
=fla +n 2 h) -fa +nh) +2h. Vfah +nh)
=fa +n -2 h)-fla +nh) +2fa + nh)-fa +n-1 h))
=fa +nh)-fa+n-1h)) - a +n-1 h)-fa +n-2 h))
= V
fa +nh) - V
fa +n -1h)
= fa+ nh)
Ag = v² (2)
fa +nh)
!22

Similarly Az =
' fa+nh)
(3) ! 73

A, =; "(n)!
fla +nh)
h"
 FINITE DIFFERENCE
OPERATORS AND
INTERPOLATION 33

Substituting the values of Ag, Aj, Az, ..A, in (1), we get

flx) =f(a +nh) + fa +nh)
(r-a+nh) + (* -a + nh)(x - a+ (n -1)h) V fa +nh)
(2) !h2
t...,

+ (* -a t nh)(x --a+ (n -
1)h)...x-a + h) V"(n)flu!h"
+ nh) ...(2)
This is
Newton-Gregory formula for backward interpolation.
Put u - ( a + nh)
,e., X= a+ nh + uh,
(where * is the value of the'variable at
which the value of function is to be
*- (a + nh) = uh interpolated)
Also
*- a+ (n-1)h =(u + 1)h

x-a+h= (u +n- 1)h.

Now putting all these values in (2), we
get
fla + nh + uh) =fla + nh) + 1 i a t nh) + u(u + 1) 2 nA
2!
Uu t )(u +)y' fa +nh)...
3!
u.(u + 1)...u+n- )'g fla +
n! nh) ...(3)
This is the form in which
ally written and this formula is Newton-Gregory formula for
useful for interpolating the valuebackward interpolatioL. is usu
tabulated values. of fx) near the end of the set of

SOLVEDEXAMPLES
Example 1. The values of f(«) = 0, 1, 2, 3....6 are given by
2 4 5 6
f): 4 17 46 97 176 289
Estimate the value off2-3) using only fiveof the given values.
Solution. Newton-Gregory formula for forward interpolation is u_eful for
fx) in the beginning of the table. Hence to find f(2-8) we finding the value of
(However we may use the valuès of fx) at x = 0 to6 or at x may use the values of fx) at x =2 to 6.
= 1to 6 etc.)
The difference table is

fx) Afe)
2 17
29
46 22
51 6
97. 28
79 6
5, 176 34
113
6 289
 NUMERICAL ANALYSIS
34

Here a= 2, h=1, x= 23
2-3 -2
= 0-3
1

By Newton-Gregory formula for forward interpolation, we have

u(2)
fa + hu) = fa) + ul) Afla) +
1! 2!
A' fa) 3f43fla) +....
(0-3)( 0-7)( 1-7) A f2)
f2:3) =f2) 4 (03) Af(2) + (0-3)(2! 0:)2A2) + 3!
1!

(0-3) . (29) + (03)( 0-7) (22) +

(0-3)(- 0-7)(- 1-7) . (6)
= 17+ 6
1! 2

= 17 + 8.7-2-31+ 0-357
= 23-747.

Example 2. Find the number of mengetting wages between Rs. 10 and Rs. 15 from the follow
ing table using Newton-Gregory formula for forward interpolation.
Wages in Rs. Frequency
0- 10.
10- 20 30
20- 30 35
30- 40 42

Solution. Newton's formula for forward interpolation is useful for finding the value off () in
the beginning of the table.
The difference table is asunder

Wages in Rs. (x) Frequency fl) Afr)

Under 10 9

30
Under 20 39 5
35 2
Under 30 74 7
42
Under 40 116

Here f(15) is required

Now a= 10, h=10,.x= 15/
15 - 10
U= =0-5
h 10
By Newton-Gregory formula for forward interpolation, we have
 FINITE DIFFERENCE OPERATORS AND INTERPOLATION
35

ull) u2) u3)

fla + hu) = fa) + Afa) + A' fla) + 3! Afa) + ....
1! 2!

f(15) =f(10) +0-51) Af10) + 0-5(2) A f(10) + 0-5(3) A f(10)

1! 2! 3!
=9+ 0-5 .(30) + (0-5)( (5) + (0-5)(- 3!
0-5X 1-5) . (2)
1
- 0-5)
2!
=9+ 15 - 0-625 +0-125= 23-6
= 24 approx.
Thus number of persons getting wages below 15
is 24.
Hence number of persons getting wages between Rs. 10 and Rs.
15 is 24-9= 15.
Example 3. Given sin 45° = 0-7071, sin 50°
=0.7660
sin 55° =- 0-8192, sin 60°=
find sin 52° by using Newton-Gregory formula for 0-8660,
forward interpolation. [M.D.U. 1998]
Solution, The difference table is as under
flx) Afx) A fe)
45 0-7071
0-0589
50 0-7660 -0-0057
0-0532 -0-0007
55 0-8192 -0-0664
0-0468
60 0-8660

Here a= 45° h=5 and s =52°

52 -45
=1:4
h " 5
By Newton-Gregory formula for forward interpolation, we have

fa+uh) =fa) + uA fa) + 2f) uu- 1)(u - 2) Afla) +

2! 3!

f(52°) =0-7071 +14x 0-0589+0x(-0-0057)

2!
+(1-4)X0-4)( -0.6)
3! x(-0-0007)
=0-7071 +0-08246 -0-001596 +0-000032 =0-7880032,
Example 4. Use Newton's formula for interpolation to find the net premium at age 25 from the
table given below :
Age Annual net premium
20 0-01427

24 0-01581

28 0-01772

32 0-01996
36 NUMERICAL. ANA:

Solution. Thedifference table is as under

Age x Premium flx) M)
20 0-01427
0-00154

24
0-00037
0-01581
-0-00004
0-00191
0-00033
28 0-01772
0-00224
32 0-01996

Here a = 20, h =4, =25

25- 20
= 125
4

By Newton-Gregory formula for forward interpolation, we have

ul) u3)
fla + hu) =fla) + 1! a fa) + 3! A fa) + ...
(1-25)(0-25) x (0-00037)
f(25) = 0-01427 + (1-25(0-00154) + 2!
(1-25)(0-25)(0-75)
+
3! x(-0-00004)
=0-01427 +0-001925 +0-0000578 +0-0000016

= 0-01625 (nearly)
Example 5. Given
6 7

fx) : 27 64 125 216 343 512

Find the value of f(7-5).

Solution. Here the value to the interpolated lies at the end of the given observations Le., near
8. Hence in thiscase Newton-Gregory formula for backward interpolation will be more appropriate.
Here a + nh=8; h=1

-a+ nh
and U=
h
7-5-8
=-0-5.
1

Now to calculate Vfla + nh), Vfa+nh)...., prepare the following difference table.
FINITEDIFFERENCE OPERATORS AND INTERPOLATION 37

The difference table is

f(x) Vfx)
1 1
fx)
7
2
12
19 6
3 27 18
37 6
4 64 24
61 6
5 125 30
91 6
6 216 36
127 6
7 343 42
169
512

Here V fr), vfx)... are all zero.

Now Newton-Gregory formula for backward interpolation is
fla + nh + uh) =fa + nh) + Vfla +nh) + u(u +1)2 ² fia +nh)
1! 2!

ulu + 1)u + 2) 3 f(a+ nh) +

3! ...
f(75)=512 + (-0-5) x 169-405)(-0:5 +1).0
2!

(-0-5)\- 0-5 +lX -0-5 +2).

x6
3!
=512 - 84-5 -5-25 -0-375 =421-875.
3
Example 6. Find the value of anannuity at 5%, given the following table :
Rate per cent Annuity value
4 17-29203

16-28889

5 15-37245

5 14-53375
6 13-76483
38 NUMERICAL ANALYSIS
Solution. Here the value to be interpolated lies near the end of tabulated values. Hence we
use Newton-Gregory backward formula for interpolation which 1s
fa + nh + uh) = fa + nh) + Vfla +nh)
1!
1)(u + 2) v' fla+ nh) +....
u(u + 1) y'f(a +nh) + u(u +3! ...(1)
2!

Here 1
a + nh=6: h=
2

-6
and uatnh
1 4

The difference table is

10°y 10°Vy 10°Vy 10°9y 10°7y

4 1729203
- 100314

1628889 8670

91644 -896
1537245 7774 100

-83870 -796
1
5 1453375 6978

- 76892
6 1376483

Substituting the values of u, fla +nh) and its differences in equation (1), we get

= 1376483 76892 +
(--)
2!
x 6978+
x(-796)
3!
5 1
4 4 4 4
x 100
4!
= 1376483 + 96115 + 1090 - 31 + 1-7=
1473658-7
1473658-7
= 14-73659
105
FINITE DIFFERENCE
OPERATORS AND INTERPOLATION 39

Example 7. If p, 9, r, s be the successive entrios

table, show that when third
differences are taken into
corrosnonding eguidlistant ague to
monts hafway between the account, the entry corresponding to the irgu-
and r, and Bis the
arguments q are ris A+ 24 B, where A is the arithmetic mean between 'q
arithmetic mean of 3q - 2p - s and 3r - 2s - p. M.D.U. 2001|
Solution. Let a be the first term and h be the
Here
fla)=p, fla +h) =q, fla + 2h) =r,interval of differencing.
fa + 3h) =s
We want to find the value 3
offla +h
3 -a
a+h
ua 3

Now applying Newton-Gregory formula for

forward interpolation, we have
3 (3
1 -1 -2
2 2|2 2
1! Afa) + 2! A' fa) + 3!
A³ fa)
3 fa) +3 fla) -
-fa) +,A 1
16
The difference table is

f) Afe) fx)
p

q-p
a+h
r-2q +p
r-q s - 3r -3q -p
a+ 2h s - 2r +q
S-r

a+35

Using the values of f(a), Af(a),.. in (1), we get

3 1
f<a+-p*a-)-4 +p)-: 16
(s - 3r +3q-p)
9 9 1
16P+ 169+ 16 16

..(2)
)
Now A=9tr
2 ...(3)

and B= (3q - 2p -s) +(3r - 2s -p)

2

Or B-3g +r-p-s)
2 .(4)
 NUMERICAL ANALYSIS
40

Using (3)and (4), equation (2) becomes

1 1 B.
3 24
16 EXERCISE 2(a)
Newton-Gregory formula for forward
1. () What do you understand by interpolation ? Obtain the
interpolation. when it is useful
interpolation. Alsostate
backward
(ü) State and prove Newton-Gregory formula 1for
to apply.
polynomialfor the following data:
2. Construct Newton's forward interpolation 10
6
16
fx): 3

Hence evaluate f (5).

3. The values off*) for x = 0, 1, 2, ...6 are given by
5 6
1 2 4
24 38
10 16 20
fl: 4

Estimate the value off(3-2) using only four of the given values.
obtained marks less than 45.
4 From the following table find the number of students who
Marks No. of Students
30 40 3I

40- 50 42

50 -60 51
60 -70 35
70 - 80 31
[K.U. 2006]
The following table gives the population of a town during the last six censuses. Estimate using any
suitable interpolation formula, the increase in population during the period from 1946 to 1948.
Year: 1911 1921 1931 1941 1951 1961
Population (in thousands): 12 15 20 27 39 52
6. The population of atown is as follows:
Year: 1921 1931 1941 1951 1961 1971
Population (in thousands): 20 24 29 36 46 51
Estimate the increase in population during the period 1955 to 1961.
Given
40 45 50 55 60 65
logx: 1-60206 1-65321 1-69897 1-74036 1-77815 1-81291
Findthe value of log 5875.
 DIFFERENCE
FINITE OPERATORS AND INTERPOLATION 41

The population of a country is as follows:

8.
Year 1891 1901 1911 1921 1931
Population (in Lakhs) 46 66 81 93 101
Bstimate the population for the year 1925,
Ifl, represents the numbers living at age x in a life table fnd accurately as data will
9.
walues ofx= 35. 42 and 47, given Lon =512, Lan = 439, permit. Iy for
L=346 and Lso =243.
10. The following are the number of
deaths in four successive ten years age groups. Find the number of
deaths at 45 - 50and 50 - 55
Age Group: 25-35 35 45 45 - 55 55- 65
Deaths: 13229 18139 24225 31496
ANSWERS
2. 1-625 3. 17-28 4. 48
B. f(1946) = 34-5l5; f(1948) = 35-903, increase in population 1-388 thousands (app.)
6. 6210 7. 3-7690
8 96-7248 10. 11278, 12947
9. 394,326,274
2.10. Central Difference Interpolation Formulae
Newton's formulae which we have discussed in the previous section are fundamental and appli
cable to nearly all cases of interpolation, but they do not converge as rapidly as the central difference
formulae. The central difference interpolation formulae are in particular suitable when it is required
to find the value of a tabulated function near the middle of the table.
The most commonly used central difference interpolation formulae are:
(i) Gauss Forward Interpolation Formula
(iü) Gauss Backward Interpolation Formula
(üi) Sterling Formula
(iv) Bessel's Formula
(v) Laplace-Everett's Formula
We will discuss each of these methods with the assumption that values of fx) at equidistant
points x1 J2, ...., *, are given.
2.11Gauss Forward Interpolation Formula (K.U. 2000; M.D.U. 1999]
We know that the Newton's advancing formula is
...(1)

where x(x - 1)(x -2),..(x -i+1) .(2)*

Now, except the first two terms, the subscript ofy in each term is diminished by 1, using the
formula
A"tly-1= yo - A"y -1
s'y, = 'y-1+A+ly-1 ...(3)

x() - XX -1) (*- 2) ...*-i+ 1) and "C, nn- 1) ...(n -r+1)

i! r!

In "C,, nis a positive integer whereas in x(), xmay not be an integer.

 NUMERICAL ANALYSS
42

Puttingn = 2, 3, 4, in thisreenlt, we get

Using these results, egn. (1) becomes )+

3, 4,.... (Compare: "C,-1 t "C, = " CJ

But å-)+T) =(t + l),where i
t(x+ 1)40'y -1t.....
V, 5Yo +ay, t x24'y +(r+ 1),g4y
is reduced by 1.
Again, the subscript ofy of ffh term and onward -2) t ...
=Vo + )Ao + g,4"y- 1 t (* + 1),a A'y- + (* + 1)4 (A'y-2+ A°y
V*
A'y-1=A'y -2 +4'y-l
[: 4°y- = A'y- - Ay-, :
This is called the GaussForward Formula.
Note. Gauss Forward formula is useful when lies between 0 and

2.12. Gauss Backward Formula

We know that Newton's advancing formula is
.... ...(1)
Y.=Yo +1) Ayo +2, 4y% + (3) Ayo +4) Ayo t
term is reduced by 1, using the formula
Here, except the first term, the subscript ofy in each
A'yc = A"y-1 +4"*y-1 ...(2)

Putting n=1,2, 3, 4,... in (2), we have

Ayo = Ay-1+Ay-1
A'y,= A'y-+ A'y-1
A'y, =A'y.1+A'y-1
Atyo =A'y-1t 4y -1

Using these results eqn. (1l) becomes

y =Yo +*a, (Ay-j +Ay-) +x (4'y- +4y-)
+*g)(a'y-1+A'y-) +4, (A'y-1tAy-)t.
=yo +*y,Ay-j+() t *e)'y-1+(Wa t *g)4'y -1 +(*9) t*4)4y-1 t ...
=yo +XaAy-1 t(*+1)aA'y- +(* +1),g,4'y -1 +(* +1)44'y-1t.. ...(3)

{"-) t )= (* + l)e
FINITE DIFFERENCE
OPERATORS AND INTERIrOLATION 43

Again sutfix oty ot all the terms of (3) excent frst t roe io rorluced by 1,
using the fornuie
A'y-1 =N'y2 + A"ly -2
Putting n = 3,4, ..., we have

A'y-= y_tA'y-2
A'y-= A'y-2+ A°y -2
Using these results eqn. (3) becomes
Y=Yo +xAy-1 +(x +1)g4'y+ (*+ 1)a [A'y -2+ A'y -
+(« +1)4, (A'y-2tAy-l+...
=Yo +aAy -1+(*+1)e4'y-+(*+ )g4y-2+ l(x + 1)9 +(* +1)4}A'y -2t *
Yo t*Ay-1+(*+1e4'y-+(+ 1)g^y- +(*+2)4,A'y-2t. .
which is the Gauss Backward Formula.
1
Note. Gauss Backward formula is useful when xlies between -and 0.
2

Table For Writing Gauss Forward and Backward Formulae :

Coeff. 1 (x +1)e) (x + 1)) (x + 2)4)

Ay -1 Ay -2
A
Ny-1 Ay -2 A

Ayo

Coeff. 1 (1) R(2) (x + 1)ø) (x+ 1)4)

For writing Gauss forward formula, yo and differences along AA' and below it are multiplied
by the corresponding co-efficients given at the bottom and the products so obtained are added and
equated to y,.
Hence Y.=yo t *u, Ayo +*, 'y -1+(*+ 1)g, 4'y -1 + (*+ 1)a A'y-t..
For writing Gauss backward formula, yo and differences along AA' and above line AA' are
multiplied by the corresponding co-eficients given at the top and the products so obtained are added
and equated to y,.
Hence y, =yo +*1) Ay -1+ (*+1), Ay -1+(*+1)3 A'y -2+ (* +2)4 A'y-2t.
44 NUMERICAL ANALYSIS

2.13. Sterling Formula |K.U. 2000|

The Sterling formula is the average of Gauss's forward and backward formula and is useful
when x lies between - 1/2 and 1/2.
We know that Gauss's forward formula is

y,=Yo +x Ay,+ x(* - 1)

2! Ny-1t (x+ 1) x( - ) y1
3!
(x+ 1)x(x- 1)(* = y9t .... ...(1)
4!
and Gauss's backward fornmula is

Vx=Yo +x A y21
+2, , ( +1) x(* - 1)
-1+
3! A'y -2
(x+ 2)(x+ l)x(*- )Ay9+ ... ...2)
4!
Taking mean of (1)and (2), we get
1 '*
Y=Yo t; lAyo +Ay-jl+ l(*-1) +(*+ 1)) Ay -1
1 (*+1) . *(* - 1)
+2
3!
IAy-1+ y-l
+
1 (x+1). x(*- 1) [(x-2) + (x + 2)] Ay-2t ...
2 4!

Or
Y=Yo +*. (Ayo +Ay-)+ 4y-1
2-1) 1 (4'y-1+4By_g)+-) Ay-2t... ...3)
3! 4!
which is the Sterling Formula.
Note. Sterling formula is useful when x lies between

Table for Writing Sterling formula :

Coeff. 1 x(g-1) -1)

2! 3! 4!

Ay -1 Ay -2

Sy -1 y -2
Ayo N'y -1
Here mean of the odd differences is to be taken but y, and even diferences are to be kept as
they are.
Arithmetic means of first and third differences in second and fourth columns are multiplied by
t and r?
3!
respectively.
FINITE DIFFERENCE
OPERATORS AND INTERPOLATION 45

Also Yo, second, fourth respectively. These pro

ducts are added and equated to y..
differences are multiplied by 1, , 4!

2.14 Bessel's Formula [M.D.U. 2000)

besselstormula
the origin can be obtained by taking the mean ofGauss backward formula after shifting
to l and Gauss forward formula.
Gauss backward formula is
)
Y=Yo + (x+1)2) (+1)3) A'y-+(+2)4)-Ay-2t... ...(1)
1! Ay-1+ 2! -Ay-1+ 3! 4!

On shitting origin to 1(this can be done by replacing xby x- 1and adding 1to each subscript
of y), we get
(*-1)) 2) 4y., 2+1)4) ...(2)
Y,=y) +
1! Ayo + 2 4%+ 4! A'y-1t...
3!
Gauss's forward formula is
x) (+ 1)8) ...(3)
y,=Yo t 1! Ayo + 2! A'y-1+ 3! 4! Aty-2t *..
Taking mean of (2) and (3), we get
1
O1ty) + 2) 1 1
)+ (x + 1)°]Ay-1
2!2
(z+ 1)4)
4!

x(*-1).2,1 LAy, +Ay-J

2! 2

-1)
..(4)
3! 4!
which is the Bessel's interpolation formula.
1
Putting z=*-ie., x =2+ we obtain the simpler and more symmetric form i.e.,

Y,=,b tyd +z yo + 2!
2.2 1) 1 9
4 4 4 1
+
3! Ay-1t 4! ;(A'y-1+4y-l t... ..5)
Note. Bessel's formula should be used whenx lies between 2
and 1 and forx =1 ,terms containing
1
odd differences vanish and thus the formula is most suitableatx=;
 NUMERICAL ANALYSIS
48

Table for Laplace Everett's Formula :

y-1
B'
B...... Yo N'y-1
Ayo A'y1
A'
A.... A'y, A'y1

Coefficients of the differences along the lines AA' are

(x + 1)8) (r+ 2)6)
1!' 3! 5!

and the coeficients of the differencesalong the line BB are

u (u+ 1)8) (u + 2)(6)
1!' 3! 5!.
where u=1-x.
Thus the Everett's formula can be written as
(x+ 1)3) (*+2)6)
3! A'y + 5! a'y-1t..
u)
(u +2)5)
+ Ay-1+ 5!
a'y-,t...
3!

2.16. Steps in Applying Various Central Difference Formulae

at one of the given values
1. For interpolating in the middle of the table, first shift the originbe
is to intérpolated. Ifa' is the
of the arguments nearer to the argument whose entry value
origin and h is the interval of differencing, then the changed values of the independent
variable are given by u = h

2. Finducorresponding to the value ofx to be interpolated.

3. Prepare a difference table for the given set ofvalues of uand the corresponding functional
values.
4 Find the limit in which the value ofx to be interpolated lies about origin and then choose a
suitable central difference formula to find the value of the function.
SOLVED EXAMPLES

Example 1. Use Gauss's forward formula to find the value ofy when x=375 from the follou
ingtable:
2-5 3-0 3-5)) 4-0 4-5 5.0
24-145 22-043 20-225 18-644 17:262 16-047
EINITE DIFFERENCE OPERATORS AND INTERPOLATION 49

Solution. We know that Gauss's forward formula is useful when xlies between 0 andHere
2
intervalof differencing is 05 and we want to find the value ofy at x=375. Therefore, shifting the
origin at 35and taking 05 as the unit i.e.,we define u= X-3-50
0-5
Now the value of yrequired will
be the value for u where u= 3-75-3.-50 = 0-5
0-5
The difference table is

*-3-5
u=
0-5 Ayu N'y
.5-2 24-145
2-102
-1' 22-043 0-284
1-818 --047
S 20-225 0-237 -009
A A
1-581 --038
18-644 0-199 -006
-1-382 -032
2 17-262 0-167
- 1-215

16-047

Gauss' Forward Formula is

u(u-1)
yu=Yo t udyo + u.(u -1) 4y+ u+ 1)3!
2!: Ay-1+ (u +1) uu- 1)(u -2) 4! Ay-2t*..
Yos=20:225 +0-5 x(- 1-581l) + (0-5)(0-5 -1) x (0-237)
2!
+ (0-5 +1\0-5)(0-5 - 1) x(- 0-038) + (0-5 +1)X0-5)(0-5 - 1X0-5 -2) x (0-009)
3! 4!
= 20-225-0-7905 -0-029625 +0-00238 +0-0002109
= 19-407 approx.
Example 2. Given that V12500 =l11-803399, V12510 = 111848111
V12520 = l11-892806, V12530= 111-937483
Show by Gauss backward formula that V12516 =111-874930.
Solution. We know that Gauss's backward formula is useful when xlies between -1/2 and
0. Here interval of differencing is 10 and we want to find the value of y at x* 12516. Therefore,
shifting the origin at 12520 and h= 10 as the unit i.e., we define
-X-12520
10
50 NUMERICAL ANALYSIS
Now the value of yrequired is for
12516 - 12520 =-0.4
10

The difference table is

Ay, Ay, A'y,

- 2 111-803399
0-044712
- 0-000017
-1 111-848111
0-000001
0-044695

111-892806
0-000018 4eesee.

...0...
0-044677
1 111-937483

Gauss backward formula is

Vu=Yo + W Ay -1+ (u + 1)2 y-1+ (u + 1)9, A'y -2t ..
Putting u=-0-4and substituting the values of y, and its differences, we get

y-0-4= 111-892806 + (- 0-4). (0-044695) +

(06)(-
2
0.4)
(0-6)( - 0-4)( - 1-4)
x(-0-000018) + 6 .(-000001)
= 111-892806 0-017878 +0-000002 = 111-874930.
Example 3. Use Sterling formul to find yzs given
Y20 =49225, ys =48316, y0 =47236, Yas =45926, y4) =44306.
Solution. Taking x=30as the origin and h =5 as the unit, we have to find the value ofy for
28 -30
U= =-0-4
5
The difference table is
t-30
5
Yu Ayu
20, -2 49225
- 909
25 -1 48316 - 171
1080. 59
A .30... ...0... 47236 230 21 .A'
1310 -80
35 1 45926 310
1620
40 2 44306
FINITE DIFFERENCE OPERATORS AND INIERIOLATION 51

Sterling formula is
Ayo + Ay 1
u(u'- 1) A'y+A'y-2, u'u'- 1)
y, =Yo tu. A'y.
2 21 3-!
Putting u=-0-4, and the values of various differences from the table, we get
2 4! y-2
y- 04 = 47236 + (-0.4) -1310 -1080 +(0.16) (- 230)
2 2

+ (-0-4)(0-16 - 1) (- 80 - 59)
6
(0-16)(0-16 - 1)(-21)
2 24
= 47236 +478 - 18-4- 3-8920 + 1176
i.e., V2s = 47692,
Example 4. Use Sterling's formula to find the value of N22-2,given that :
V20 =4-472, N21 =4-583, V22 =4-690, V23 = 4-796, N24 =4-899.
Solution, Take the origin at 22 and h = 1 as unit.
x-22
Let u=
1

We have to find value ofy for u = 22-2- 22-0 = 0-2

The difference table is

Ayu Ay A'
-2 4-472
0-111
-1 .4-583 -0-004
0-107 0-003
A ...0... 4-690 0-0017 0-005 A':
0-106 0-002
1 4.796 -0-003
0-103
2 4-899

Sterlings formula is y,=yo +u. (Ayo +Ay-i)+ u' y-1

u(u²-1) 1
3! (a'y-1+Ay-)+ u'(u'-
4!
1) Ay-2t...

Yo2= 4690 +02 x 2 (0-107 +0-106) +(0.2)2

2! x(-0-001)
(02X0-04-1) 1 (0-003 - 0.002)+009/004
6 24
x-0005) +....
= 4-690 +0-0213-0-00002 =4.71128
i.e., Y22,2 =4-71128.