Chapter 1: Basic Concepts in Error Estimation

1.1. Introduction

Most phenomena can not solved by analytical (exact) methods,

led to evolve numerical methods. The aim of numerical analysis
is to provide constructive methods for obtaining answers to
such problems in a numerical form.
Definition
Numerical Analysis is the area of mathematics and computer
science that creates analyzes and implements algorithms for
solving problem numerically. In short it designs (constructs) of
algorithms

DBU Page 1
Cont. . . .

The 3 central concepts in numerical analysis are:

i. Convergence: Whether the method approximates the
solution
ii. Order: How well it approximates the solution
iii. Stability: Whether errors are damped out.

Definition
Numerical Methods emphasize on the implementation of
algorithms. Therefore, the aim of numerical methods provide
systematic methods for solving problems in numerical form. In
short it implements of algorithms.

DBU Page 2
Cont. . . .

Numerical methods require tedious and repetitive arithmetic

operations to get approximate solutions of problems with the
desired accuracy. Due to this, the course requires Computer
Software. Most of the numerical methods will be described fully
through pseudo-code in such a form that translating this code
into computer programs:
[MATLAB, C, C + +, Mathematica, fortran, python] is relatively
straightforward. In this course, we use MATLAB computer
language.

DBU Page 3
1.2. Source of errors
a. Error of the problem (Modeling error): Mathematical
modeling is the process when mathematical equations are
used to represent a physical system (phenomena). But
mathematical modeling rarely gives an exact picture of
actual phenomena the most parts are idealized models.
So, in showing the phenomena of nature are forced as a
rule to accept certain conditions.
Example: Moving a spring from one end to another without
air resistance.
b. Blunders and mistakes : Blunders occur at any stage of
the mathematical modeling process. Such error can be
avoided by taking proper care in approach and design to a
solution.
Example: Mistakes due to the programming error.

DBU Page 4
Cont. . . .

c. Machine representation and Arithmetic error: These

errors are inevitable when using floating-point arithmetic
when using computers or calculators
Example: Rounding and chopping decimals.
d. Mathematical approximation error: These errors arise
from use of finite formula instead of infinite formula.
Let x be the exact value of a quantity and x ? be its
approximate value. Then
error = x − x ?
Exact value is a√number which no uncertainly is associated.
Example 5 , 21
3 , 3 etc.
Approximate value
√ is the number obtained by retains a few
digits. Example 3 = 1.732, π = 3.142.
DBU Page 5
1. Significant digits (figures) are the number of digits used to
express the number. For instance, each of the number
1.360, 1360, 0.01360 has four significant digits. But
sometimes, the number 1360 may have three or four
significant digits depending on whether zero is known with
confidence. Such uncertainty can be resolved by using
scientific notation. That is 1.36 × 103 has three significant
digits but 1.360 × 103 has four significant digits.
2. Accuracy refers to how closely a computed(measured)
value agrees with the exact(truth) value.
3. Precision refers to how closely a computed(measured)
value agrees with each other.

DBU Page 6
1.3. Types of errors
a. Inherent error: Errors which are already present in the
statement of a problem before its solution. These type of
error arise either due to the given data being approximate
or due to the assumptions made in the mathematical
modeling of problem. Inherent error can be minimized by
taking better data.
b. Rounding (Round off) error: arise from the process of
rounding off the numbers during the computation.
Note: There are rule to round off a number to n significant
figures.
c. Truncation error: arise from replacing an infinite process
by a finite one.
2 3 4
Consider: ex = 1 + x + x2! + x3! + x4! + · · ·
This formula is infinite, we can’t calculate ex at any x.
DBU Page 7
Cont. . .

x2
Take the first few terms. If you take ex ≈ 1 + x + 2! , then
2 3 4
Exact value = 1 + x + x2! + x3! + x4! + · · ·
2
Approximate value = 1 + x + x2! and
3 4
error = x3! + x4! + · · ·

Measure of Errors
Numerical solutions (approximate solutions) are subject to
certain errors.
1 Absolute error (Ea ) = |error | = |x − x ? |
Ea
2 Relative error (Er ) = |x|
3 Percentage error (Ep ) = 100 × Er

Note: Er and Ep are unit-less whereas unit of Ea is the unit of x.

DBU Page 8
Example
Find absolute error, relative error and percentage error if
x = 2.538 is rounded-off
a. To one decimal places
b. To three significant figures (digits)
c. Chopping to two decimal places

Solution

a. Exact value(x) = 2.538 and approximate

value(x ∗ ) = 2.5.
Error= x − x ? == 2.538 − 2.5 = 0.038
Ea = |error | = 0.038
Ea 0.038
Er = |x| = 2.358 = 0.01611535199
Ep = 100 × Er = 100 × 01611535199 = 1.611535199
DBU Page 9
cont...

b. Exact value(x) = 2.538 and approximate

value(x ∗ ) = 2.54.
Error=exact value—approximate value
= 2.538 − 2.54 = 0.002
Ea = |error | = 0.002
Ea
Er = |x| = 0.002/2.358 = 0.00084817642.
Ep = 100 × Er = 100 × 0.00084817442 = 0.84817642.
c. Exact value(x) = 2.538 and approximate
value(x ∗ ) = 2.53.
Error=exact value—approximate value
= 2.538 − 2.53 = 0.008
Ea = ..............
Er = .............
Ep = ................
DBU Page 10
The absolute error refers to the number of digits after the
decimal point.
Theorem
If a number be rounded to n decimal places, then
Ea ≤ 12 × 10−n

Example
Find the number of significant figures in the number 6.8315 if its
relative error is 0.1 × 10−3

Solution

Ea = exact value × Er = (6.8315)(0.1 × 10−3 ) ≈ 0.7 × 10−3

=⇒ 0.7 × 10−3 = 0.07 × 10−2 ≤ 12 × 10−2
=⇒ The number is correct to two decimal places.
Hence, the number of significant figure is three. i.e. 6.83
DBU Page 11
1.4. Propagation of errors
1.4.1. Errors in the approximation of a function

Consider a function u = f (x1 , x2 , · · · , xn ) where x1 , x2 , · · · , xn

are variables. Let the error in each xi be ∆xi for (i = 1, 2, ..., n).
Then the error ∆u in u is given by

u + ∆u = f (x1 + ∆x1 , x2 + ∆x2 , · · · , xn + ∆xn )

Apply,Taylor’s series expansion for function of several variables,

we get
∂u ∂u ∂u
u + ∆u = f (x1 , x2 , · · · , xn ) + (∆x1 ∂x1
+ ∆x2 ∂x2
+ · · · + ∆xn ∂xn
)+
1 2 ∂2u 2 2
+ (∆x2 )2 ∂∂xu2 + · · · + (∆xn )2 ∂∂xu2 + · · · ] + · · ·
2 [(∆x1 ) ∂x 2 n
1 2
Assume the errors ∆x1 , ∆x2 , · · · , ∆xn in xi are small and that
∆xi
xi < 1.

DBU Page 12
So, the terms containing (∆x1 )2 , (∆x2 )2 , · · · , (∆xn )2 and higher
powers are neglected
∂u ∂u ∂u
u + ∆u = f (x1 , x2 , · · · , xn ) + (∆x1 ∂x1
+ ∆x2 ∂x2
+ · · · + ∆xn ∂xn
)
∂u ∂u ∂u
Therefore, ∆u = ∆x1 ∂x1 + ∆x2 ∂x2 + · · · + ∆xn ∂xn is error in the
function u .
absolute error = |∆u|
Ea
relative error= |u|
∂u ∂u ∂u
maximum absolute error= |∆x1 ∂x1
| + |∆x2 ∂x 2
| + · · · + |∆xn ∂x n
|
maximumEa
maximum relative error= |u|

Example
2
Let f (x, y , z) = 5xy
z3
. Find absolute error and relative error if
0.1, −0.1 and 0.2 are errors in x, y and z at (x, y , z) = (1, 1, 1)
respectively?

DBU Page 13
Solution

∂f 5y 2 ∂f 10xy ∂f −15xy 2
∂x = ,
z 3 ∂y
= z3
and ∂z = z4
∂f ∂f ∂f
∂x (1, 1, 1) = 5, ∂y (1, 1, 1) = 10 and ∂z (1, 1, 1) = −15
∂f ∂f ∂f
∆u = ∆x ∂x (1, 1, 1) + ∆y ∂y (1, 1, 1) + ∆z ∂z (1, 1, 1) =
(5)(0.1) + (10)(−0.1) + (−15)(0.2) = −3.5
Ea 3.5
Ea = | − 3.5| = 3.5 and Er = |f (1,1,1)| = 5 = 0.7

1.4.2. Error in operation of numbers

a. Error in addition of numbers
Let u = x1 + x2 + · · · + xn , where x1 , x2 , · · · , , xn are
numbers. Then
u + ∆u = x1 + ∆x1 + x2 + ∆x2 + · · · + xn + ∆xn
Therefore, ∆u = ∆x1 + ∆x2 + · · · + ∆xn
DBU Page 14
 b. Error in subtraction of numbers
Let u = x1 − x2 . Then
u + ∆u = x1 + ∆x1 − (x2 + ∆x2 )
⇒ ∆u = ∆x1 − ∆x2
c. Error in product of numbers
Let u = x1 × x2 × · · · × xn be number. Then the error in u is
∂u ∂u ∂u
∆u = ∆x1 ∂x 1
+ ∆x2 ∂x 2
+ · · · + ∆xn ∂xn
d. Error in Division of numbers
Let u = xx21 be number. Then error in u is
∂u ∂u
∆u = ∆x1 ∂x 1
+ ∆x2 ∂x2

DBU Page 15
Exercises
1. Find Ea and Er in the function f (x) = 3x 7 − 6x if 0.05 is the
error in x at x = 1?
2. Given x = 30.25, y = 29.97 and z = 14.75
Find Ea and Er in the numbers x + y + z , xy − z, xyz and
z
xy if each the given numbers rounded to three significant
figures?
3. Determine the number of correct digits in the number
841.256 if its absolute error is 0.1?
4. The dimensions of a rectangular field are measured to be
40m and 60m. Find Ea and Er in the area of the field if
0.1m is Ea in each dimension? what will be the exact area
of the field?

DBU Page 16