Fractional Relative Error
Uploaded by
ScribdTranslationsFractional Relative Error
Uploaded by
ScribdTranslationsRelative fractional error = error / true value
Where:
Error = true value - approximate value.
The relative error can also be multiplied by 100% to express it as Ev = (error
true/ true value ) 100; Where Ev denotes the relative percentage error. The subscript v
it means the normalization of the error to the true value.
For numerical methods, the true value will only be known when it is referred to
functions that can be solved analytically. However, in real applications, it is not known
the true answer. In these cases, normalizing the error is an alternative using the best
possible estimation of the true value, this is to the approximation itself, such as:
Ea = (approximate error / approximate value)100
Where the subscript a means that the error is normalized to an approximate value.
One of the challenges faced by numerical methods is to determine estimates of
error in the absence of knowledge of the true values. The error is calculated as the difference
between the previous approach and the current one. Therefore, the percentage relative error is given by
Ea = abs(((current approximation - previous approximation) / current approximation) * 100)
If the previous relationship is fulfilled, then it is considered that the obtained result is within the
acceptable level, that is, to a previously set error (Es):
Abs(Ea) <>
ROUNDING ERRORS
Rounding errors occur because computers only store a finite number of digits.
significant during a calculation. Computers perform this function in different ways;
this technique of retaining only the first seven terms was called 'truncation' in the environment of
computation. It will preferably be called truncation, to distinguish it from rounding errors.
A cut ignores the remaining terms of the complete decimal representation.
Most computers have between 7 and 14 significant digits, rounding errors
they might not seem very important. However, there are two reasons why they can be significant.
critics of some numerical methods:
Certain methods require extremely large amounts to obtain an answer.
In addition, these calculations often depend on each other, that is, subsequent calculations are
dependent on the previous ones. Consequently, although an individual rounding error may
being very small, the accumulation effect over the course of the large number of calculations can
be significant.
The rounding effect can be exaggerated when algebraic operations are performed that
they use very small and very large numbers at the same time. Since this case occurs in
In many numerical methods, rounding error can be very important.
TRUNCATION ERRORS
Truncation errors are those that result from using an approximation instead of an
exact mathematical procedure.
These types of errors are evaluated with a mathematical formulation: the Taylor series.
Taylor is a formulation to predict the value of the function at Xi+1 in terms of the function and
of its derivatives in a neighborhood of the point Xi.
Being the final term:
Rn = ((ƒ(n+1) (ξ))/(n+1)!)hn+1
In general, the Taylor series expansion of n-th order is exact for an n-th degree polynomial.
order. For other continuously differentiable functions, such as exponentials or sinusoidal functions, they do not
gets an exact estimate through a finite number of terms. Each of the terms
Additional factors contribute to the improvement of the approach, even if it is just a little.
TOTAL NUMERIC ERROR
The total numerical error is the sum of the rounding errors and truncation errors. The only way to
minimizing rounding errors involves increasing the number of significant figures of the
computer.
Graphic representation of the advantages and disadvantages between rounding errors and truncation that
sometimes influence the course of a numerical method. The optimal point shows where the error
Rounding begins to negate the benefits provided by the reduction in step size.
ERRORS BY MISTAKE
In the early years of computing, erroneous numerical results were attributed
sometimes due to the malfunctioning of the computer itself. Nowadays, this source of error is very
improbable and most of the mistakes are attributed to human errors.
Mistakes occur at any level of the mathematical modeling process and can
contribute with all the other components of the error. Mistakes generally go unnoticed.
high in the discussion of the numerical method. This undoubtedly proves the fact that errors of
clumsiness is, to some extent, inevitable.
FORMULATION ERRORS
Formulation or modeling errors lead to what could be considered a
incomplete mathematical model, since if a deficient model is being used, no method
numeric will generate the appropriate results.
