Error analysis in numerical analysis
the introduction
Numerical analysis is an important branch of mathematics, linking analytical mathematics
with computer science. It is usually used to find solutions to problems that cannot be
solved using analytical mathematics. The result we obtain is an approximation. Since we
obtain an approximate result or an approximate solution, this means that there is a
percentage of error that we must calculate. However, if we can find the error, we can find
the actual (true) solution, which means that finding the error is impossible. Therefore, we
seek to find an approximation of the error or the magnitude of the error (the value that the
error does not exceed). Thus, the task of numerical analysis is summarized in finding the
.approximate solution to a problem and evaluating the line
Approximation
Most of the numbers we deal with are approximate numbers, because they often represent
lengths, measurements, or values of physical quantities resulting from a measurement, and
are themselves approximate. Also, many real numbers cannot be expressed with a finite
number of digits. For example, the number It is approximately equal to 3.14159, and the
same is true for the number e. For example, we cannot write them as exact numbers. As
we mentioned previously, we must use approximation to solve unsolvable problems in
analytical mathematics, but this approximation will result in errors, the most important of
:which are the following
Rounding
One of the most important sources of errors is the use of rounded numbers instead of
exact ones. The rounding rule: If we have the number We wanted
to be satisfied with (n-1) the number to the right of the decimal point, meaning that the
:rounded number would be as follows
Then
It is defined as follows: If (greater than or equal to 5), then if it was
(less than 5), then For example, rounding the numbers 0.045651 and
.9033 to three decimal places yields 0.0457 and 9033, respectively
The origin of the numerical error here is due to the sufficiency of a certain number of
.decimal places after the decimal point in the calculations
chopping
Electronic calculators often do not round numbers, but rather truncate them. The
truncation rule: If we have the number If we want to be
satisfied with (1-n) the number to the right of the decimal point, then the discrete number
is as follows: For example, rounding to three
.decimal places gives 0.045651 and 9033 0.0456 and 9030 respectively
Note: The number 0 to the right of the decimal point is not considered in the
.rounding process
The numerical error here comes from being satisfied with a certain number of terms of the
.infinite series
Errors
Errors play a pivotal role in numerical analysis and determine the accuracy and speed of
:the method used later. The following are types of errors and their impact
Absolute Error
It is the absolute error committed in the rounded number. It is the absolute value of the
difference between it and the actual value X Often
the actual number x is unknown, then the absolute error cannot be determined, so we
resort to finding an upper limit for this error such as And it satisfies the inequality:
.And so we get: That is, the actual
number lies between the two numbers. , , where the value
Represents rounding the number x by increasing and decreasing the value. It
.represents an approximation by decreasing
Example
:Let x=3.257 and let the approximate value of x be 3.26 by setting the absolute error
Relative error
The relative error R made in an approximate number It is the ratio of the absolute error of
the perpetrator of this number to the absolute value of the actual number (x≠0) X) i.e.:
In case the actual value of X is not known, we resort to the upper limit of this
:error, which is
And from it: Therefore, it can be considered: So it is: The
maximum relative error committed in the rounded number Since It can be written
Example
:Let x=3.257 and let An approximate value of x is the relative error
The relative error is used to calculate significant digits, as follows: where
t is the largest non-negative integer for which we can find the largest approximation
:interval for the actual error. General notes
It is important to note that relative error is not affected by error much, while absolute
.error is significantly affected; therefore, we mostly rely on relative error
Also, the smaller the resulting error value, the better. Conversely, the larger the error
value, the more we need to look for a mathematical procedure that works to reduce
.(improve) the resulting error value
Maximum relative error in a function of several variables
:Suppose a function f of variable n
That is: We assume that the single errors in these variables are, in
order: The total error in the function, neglecting the second-order
:limits and above, is
if
:Therefore, the maximum relative error is
Example
The largest relative error in each of the following two functions in terms of the variables is:
And he made a mistake
:the solution
It is generally observed that the actual error in a value is much less than the maximum
error; this is because errors, from a scientific point of view, tend to cancel each other out
(some being positive and some negative). For example, if 200,000 numbers are added, and
each number is rounded to 4 decimal places, the maximum error is equal to
Or equal to 1, which is a large value, while the total expected error is
.usually within 0.005
References
]3[ ]2[ ]1[
Numerical Methods and Numerical Analysis / Prof. Dr. Abu Bakr Ahmed Al-Sayed / Al-Falah .1
Library
Numerical Analysis / Prof. Mahmoud Abu Al-Ezz / Al-Rushd Library .2
Numerical analysis and its calculation method / Dr. Muhammad Mansour Subh / Al-Rashd .3
Library
