In general, for any error, the relationship between the exact number and obtained by approximation is defined as:
Error = Actual value-estimated value
Sometimes they know exactly the value of the error, denoted as Ev, or we estimate an approximate error.
Now, to define the magnitude of error, or incidence on calculating the error detected, we can normalize its value:
Ea = relative error (fraction) estimated error = true value I
Since the value of Ea can be either positive or negative, in many cases we want to know more the magnitude of the error, in which case we will use the absolute value of this.
An interesting case is an investigation conducted by Scarborough, which determined the number of significant figures contained in the error as:
If we replace it in Eq. We will get the number of significant figures is the approximate value obtained reliable.
Thus, if our calculation has an error less than the criterion for two significant figures, we get numbers that correspond to a minimum:
Es = (0.5x 102-2)% = 0.5%
TAYLOR SERIES
Special attention is the approximation of functions using Taylor series expansion. Thus, if a function is continuous and differentiable in the range of interest can be written as a finite power series, or series of Taylor.
This method can not, however, be used to fit experimental data [Xi, f (x) J], but to transform known and differentiable functions to a more user-friendly.
There are certain observations that should be known by applying this formula. For example, to get a better approximation of the function to an interval [a, b], the value of Xo must be chosen as close as possible to the center of this range. This will minimize the maximum contribution of the term (X-Xo) n + l of the residue in the calculation of R (x) between
a <= x <= b. Another way to minimize the value of the waste is to raise the degree of the polynomial adjustment, or include more terms in the series and reduce the exponent of R (x). Edited from: www.norte.uni.edu.ni/estudiantes/aproximaciones.doc