Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd
Upload
22 views

Numerical Analysis: Lecture - 3

This document discusses error analysis in numerical calculations. It provides formulas to calculate relative error and absolute error in multiplication, division, powers, and sums based on the errors in the original values. It then works through several examples applying these formulas to estimate errors and true value ranges for various calculations involving rounded numbers.

Uploaded by

Z_Jahangeer
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
22 views

Numerical Analysis: Lecture - 3

This document discusses error analysis in numerical calculations. It provides formulas to calculate relative error and absolute error in multiplication, division, powers, and sums based on the errors in the original values. It then works through several examples applying these formulas to estimate errors and true value ranges for various calculations involving rounded numbers.

Uploaded by

Z_Jahangeer
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 18

NUMERICAL ANALYSIS

Lecture -3

Muhammad Rafiq
Assistant Professor
University of Central Punjab
Lahore Pakistan

ERROR IN MULTIPLICATION

Suppose exact value of a number be= x1


Suppose exact value of another number be= x2
Suppose approximate value of a number be= x1
Suppose approximate value of another number be= x2
Error in first number:e1=x1-x1
Error in first number:e2=x2-x2
Let Z be the product of two numbers=Z=x1.x2
Let Z be the product of approximate values=Z=x1.x2
Error=ez=Z-Z

= (x1.x2 )- (x1.x2)
= x1.x2- (x1- e1)(x2-e2)
=x1x2-( x1 x2- e1x2-e2x1-e1e2)
=

x1x2-( x1x2- e1x2-e2x1)

Neglecting e1e2 as it is very small


=x1x2-x1x2+e1x2+e2x1
So
ez

=e1x2+e2x1

Absolute error=|ez|=|e1x2+e2x1||e1x2|+|e2x1|
And

=|ez||e1x2|+|e2x1|

Relative error= R.E=Absolute Error/|Z|

=|e1x2+e2x1|/|x1x2|
=| e1x2 +e2x1/x1x2 |
=| e1x2/x1x2 + e2x1/x1x2 |
R.E=| e1/x1 + e2/x2|| e1/x1|+| e2/x2|
So
R.E| e1/x1|+| e2/x2|
ERROR IN QUOTIENT
Suppose exact value of a number be= x1
Suppose exact value of another number be= x2
Suppose approximate value of 1st number be= x1
Suppose approximate value of 2nd number be= x2

Error in first number = e1=x1-x1


Error in second number=e2=x2-x2
Let Z be the quotient of two numbers=Z=x1/x2
Let Z be the quotient of approximate values: Z=x1/x2
Error=ez=Z-Z
=x1/x2-x1/x2
=x1/x2 - x1-e1/x2-e2
=

x1/x2 - x1(1-e1/x1)/x2(1-e2/x2)

= x1/x2 -x1(1-e1/x1) x2(1-e2/x2)-1


= x1/x2 - x1(1-e1/x1) x2(1- (-1 )e2/x2)
[As (1+x)n=1+nx+]

= x1/x2-x1(1-e1/x1)x2(1+e2/x2)
= x1/x2-x1x2 (1-e1/x1+e2/x2)
= x1/x2-x1/x2+e1/x2-x1/x2e2
= e1/x2 - x1/x22 e2
R.E=A.E/|Z|
| e1/x2-x1/x22. e2|/|x1/x2|
|(e1/x2-x1/x22. e2 )/ x1/x2|
|e1/x1 - e2/x2||e1/x1| + |e2/x2|
| |

| |

| |

ERROR IN POWER
Let a number be=x
Such that Z=xn
Appraximate Value of x=x

Error in x: e=x-x
Error in Z: e=Z-Z
=xn-(x)n
=xn-(x-e)n
=xn-xn(1 - e/x)n
=xn-xn(1+n(-e/x))
=xn-xn+nexn-1

So
ez = nexn-1
R.E=|nexn-1|/|xn|
=| nexn-1/xn|
=|ne/x|
=|n||e/x|
EXAMPLE # 1
Given the data

then estimate the relative error, maximum

absolute error, the range in which true answers lie.


Solution:
Let x1=4.0643

e1=1/210-4

Let x2=37.487

e=1/210-3

Z=x1/x2
= 4.0643/37.487
= 0.1084.
R.E e1/x1 + e2/x2
1/210-3/4.0643 + 1/210-3/37.487
2.5640 10-5
A.E R.E Z
0.10842.564010-5

2.77937610-6
Max. Absolute Error=2.77937610-6
The range in which true answer lies
Z-A.E Z Z+A.E
0.1084-2.779410-6Z0.1084+2.779410-6
0.108397 Z 0.108402
(Correct to 3 d.p)
EXAMPLE # 2
Given the data
the estimate the relative error ,
maximum Absolute error and the range in which answer lies.

Solution:
Z=(48.425)1/2
=6.9588.
LET x=48.425
R.E=n |

e=1/210-3
|

=1/2(110/248.425)
= 5.162610-6
Maximum Absolute error = R.E Z
=5.1626106.9588
= 3.592510-5

The range in which true answers lie:


Z-A.E Z Z+A.E
6.9588- 3.5925 10-5 Z 6.9588- 3.5925 10-5
6.9518 Z 6.9588
(Correct to 2 d.p)
EXAMPLE # 3
If given numbers are rounded estimate the relative error and
absolute error in the product 4.06430.37487
Solution:
Let x1=4.0643

e1=1/210-4

Let x2=0.37487

e2=1/210-5

Let Z= x1 x2
= 4.06430.377487
= 1.52358.
A.E : ez x1 . e2 + x2 . e1
(4.06431/210-3)+(0.374871/210-4)
3.906510-5
R.E e1/x1 + e2/x2 |

110-4 / 24.0643 + 110-5 / 20.37487


2.5640 10-5

EXAMPLE # 4
Given

find relative error, absolute error and the

range in which true answer lies.


Solution:
Z=
=

= (1.8778)1/2
Z = 1.37034
R.E1/21/210-4/1.8778

n=1/2

e=1/210-4

A.E 1.3313410-5
R.E|Z |
1.3313410-5 1.37034
1.8243810-5
Z-A.E Z Z+A.E
1.37034-1.8243810-5Z 1.37034+1.8243810-5
1.37032 Z 1.370358
(correct to 4 d.p)
EXAMPLE # 5.Let 65.43 and 17.0591 be correctly rounded to
the number of digits shown. What is the smallest interval in
which the exact sum of numbers lie.

SOL:

x1=65.43

e1=1/210-2

x2=17.0591

e2= 1/210-4

Let Z=x1+x2
=65.43+17.0591
=82.4891.
A.Eeze1+e2
510-3+510-5
510-3
R.E A.E/|Z|
510-3/|82.4891|
6.0610-5

Z-A.E Z Z+A.E
82.4891-510-3 Z 82.4891+510-3
82.4841 Z 82.4941

Exercise
Q # 1:Given the data
.Then estimate the
relative error, maximum absolute error, the range in which true
answers lies.
Q #2:Given the data

then estimate the relative error,

maximum absolute error, the range in which true answers lie.


Q # 3:Given

find relative error, absolute error

and the range in which true answer lies.


Q #4:Given

find relative error, absolute error

and the range in which true answer lies.

You might also like