Convergence Analysis of Newton-Cotes Methods:

Optimizing Sub-Intervals Selection for Precise

Integral Approximation

Lea A. Catapang1, James Ray Mendaje2 and Venus D. Siong3

1,2
School of Foundation, Bahrain Polytechnic, Bahrain
[email protected], [email protected]
3
College of Graduate Studies, Don Mariano Marcos State University, Philippines
[email protected]

ABSTRACT
This study explored the piecewise approach of the closed Newton-Cotes quadrature formulas
(Trapezoidal, Simpson’s 1/3, and 3/8 rules) and how well they work with different kinds of functions in
terms of convergence and accuracy. MATHEMATICA software was used to approximate the integrals
and determine their errors, allowing for a comparison of convergence and accuracy. Simpson’s 1/3 and
3/8 rules consistently outperformed the trapezoidal rule, demonstrating faster convergence and greater
accuracy across a wide range of functions. However, as tolerance levels increased to a considerable
magnitude, Simpson’s 3/8 rule emerged as the most robust among the three methods. We recommend
investigating various domains to substantiate the findings of this study including a comprehensive error
analysis that includes truncation error, round-off error, and error bounds to provide a more detailed
understanding of the sources and magnitude of errors and to include higher-dimensional integrals to
provide valuable insights into the robustness of these methods.

KEYWORDS
Numerical Quadrature, Newton-Cotes, Trapezoidal Rule, Simpson’s 1/3 rule, Simpson’s 3/8 rule

1. INTRODUCTION
The Newton-Cotes quadrature methods are a class of numerical integration techniques based on
polynomial interpolation over a set of equally spaced nodes [1]. These techniques provide a
direct approximation of definite integral . However, they can quickly become
unstable when applied to functions with distinct characteristics specially those that are not well-
approximated by low-degree polynomials over large intervals. This leads to a substantial
increase of error, as the polynomial interpolant may not accurately capture the behaviour of the
integrand over a large interval.
The Newton-Cotes formulas are obtained by interpolating polynomials which approximate the
tabulated function . Then integrating the function over some interval divided
into equal sub-intervals such that and [2]. The two types of Newton-
Cotes quadrature methods are open and closed. The -point closed Newton-Cotes
method includes endpoints of the closed interval as nodes. It uses nodes
[3]. Unlike
closed Newton-Cotes, open Newton-Cotes formulas do not include the endpoints of .
The most recognized closed Newton-Cotes formulas are the trapezoidal rule and Simpson’s 1/3
and 3/8 rules, mainly because of their balance of simplicity and preciseness. These methods are
widely applied across various areas of study, such as engineering, economics and finance,
computational science, and more [4]. Trapezoidal rule and Simpson’s rule are interpolatory
quadrature based on linear and quadratic interpolants, respectively [5]. These low-order rules
are often slow to converge and return inaccurate results over large intervals due to the
oscillatory nature of high-degree polynomials.
High-order quadrature techniques would be necessary to integrate high-degree polynomials,
however the values of the coefficients in these techniques are difficult to obtain and often
numerically unstable [1][3]. Alternatively, the Composite Quadrature rules for trapezoidal and
Simpson’s are simpler approach and considered an effective way to increase the accuracy of the
result when integrating high-degree polynomials. These rules involve partitioning the interval
into subintervals then using low-order interpolants (e.g. Trapezoidal and Simpson) to
approximate the integrals in each subinterval.
The composite approach in numerical quadrature, particularly, using low-order Newton-Cotes
methods like the trapezoidal rule and Simpson's rule has been extensively studied due to its
practical advantages in improving accuracy and managing the convergence of integral
approximations. The study by Udin, M.J.et al evaluated and compared the performance of
Trapezoidal, Simpson’s 1/3, and 3/8 rules in terms of accuracy and efficiency, utilizing error
analysis to determine which method performs better in providing accurate results [6]. Daan
Huybrechs, reviewed and presented the difference between the low-order variants of the class of
Newton-Cotes quadrature and the high-order quadrature (Least-Square quadrature) in terms of
numerical stability and convergence properties [1]. Yalda Qani used the advanced family of
closed Newton–Cotes numerical composite formulas to demonstrate some of the computational
capabilities of the Maple package [7]. In the A.H. Nzokem study, he implemented both
analytical and numerical methods (composite Newton-Cotes quadrature formulas) in solving the
Gamma Distribution Hazard function [8]. Sheehan Olver investigated and explored the
numerical integration of highly oscillatory functions, over both univariate and multivariate
domains, using the combination of Filon-type methods and Levin-type methods [9].
Furthermore, Hamid Mottaghi Golshan proposed a numerical iterative method based on Picard
iterations and Newton-Cotes rules that can be used to approximate the multidimensional
Fredholm–Urysohn integral equation of the second kind [10]. The study of Magalhaes, P. et al.
discussed and compared the closed and open Newton-Cotes quadrature formula, utilizing
twenty segments to assess the accuracy and convergence of the two methods [11]. In the work
of Ali, A. J. et al, they utilized error and stability analysis to evaluate the degree of accuracy of
analytical, numerical (Trapezoidal and Simpson’s 1/3 and 3/8 rules), and software-assisted
(MatLab) methods [12]. Likewise, Erme Sermutlu conducted a study utilizing MatLab and
error analysis to compare Newton-Cotes and Gauss quadrature methods over the same number
of intervals for diverse types of functions [13]. Additionally, John Roumeliotis used the
software-assisted method (Maple) to prove the Ostrowski and corrected Trapezoidal inequalities
and stated two new fourth-order quadrature formula [14]. Wu, J et. Al. explored and used the
composite Newton-Cotes methods for computation of Hadamard finite-part integral with the
second order singularity focusing on their pointwise superconvergence properties [15]. And, in
the study conducted by Clarence Burg, he presented a new closed Newton–Cotes type of
quadrature formula that uses first and higher order derivatives to increase the order of accuracy
of the numerical approximations of definite integrals [16].
This paper, like the cited studies, aims to investigate the Newton-Cotes quadrature methods
(Trapezoidal and Simpson’s rules) for approximating definite integrals and analyse their
convergence and accuracy. However, it will focus on the piecewise approach (composite rules)
of the closed Newton-Cotes quadrature methods and explore their applicability to various
special function types. It will also determine how the selection of sub-intervals affects the
accuracy and convergence of these approaches.
2. MATERIALS AND METHODS
Presented in this section is the piecewise approach of the three commonly used closed Newton-
Cotes quadrature formulas.

2.1. Closed Newton-Cotes Quadrature Formulas

Newton–Cotes methods consist of approximating the integrand by a polynomial of degree ,
which matches at evenly spaced nodes [10]. The -point closed Newton-Cotes
method includes endpoints of the closed interval as nodes [3, p.198]. It uses nodes
assuming the
form

(2.1)
where the weight

(2.2)
Using the first order Lagrange interpolating polynomial, (2.1) and (2.2) return

with,

and

Combining and simplifying, we obtain the standard trapezoidal rule,

(2.3)
Since (2.3) can also be expressed as

In a similar manner, using the second-order Lagrange interpolating polynomials, (2.1) and (2.2)
give
with,

We obtain the standard 1/3 Simpson’s rule, which is one of the most recognized closed Newton-
Cotes quadrature approaches in practice.

(2.4)

Since (2.4) can also be written as

Likewise, utilizing the third-order Lagrange interpolating polynomial (2.1) becomes,

with
Combining and simplifying, we obtain the 3/8 Simpson’s rule

(2.5)

Since (2.5) can be

expressed as

2.2. Measure of Exactness

Definition 2.1 [5, p. 121] An interpolatory quadrature rule has degree of exactness (degree of
precision) if for all (polynomial interpolant),

The definition indicates that a degree- quadrature method has a degree of exactness
This further implies that it can exactly integrate . However, there are exceptional cases
wherein a degree- interpolant can also exactly integrate higher degree polynomials.
Consequently, the degree of precision of trapezoidal rule and Simpson’s rules are one and three,
respectively.

2.3. Error Analysis for Closed Newton-Cotes Quadrature Methods

The following theorem serves as basis for error analysis of the closed Newton-Cotes quadrature
formulas since they are founded on polynomial interpolation and weighted sum of function
values at specific nodes.
Theorem 2.1 [5, p. 217] (Interpolation Error Formula) Suppose and let
denote the polynomial that interpolates { for distinct points
Then for every there exist such that

(2.6)

From this formula follows a bound for the worst error over :

Proof. Let be some arbitrary points in the interval and let be

the interpolation error of . We want an expression to describe If for any
then and choosing arbitrary in returns (2.6).
Now, suppose for all , to describe we define a function for
in [ by

Since and , it follows that For , we obtain

Furthermore,

Hence, and at the distinct numbers By Rolle’s

theorem there exist a number in for which . And the derivative of
is

Since is of degree at most , then the derivative . Additionally,

is of degree polynomial, thus
Equation (2.6.1) becomes,

Therefore, for some

Theorem 2.2 [3, p. 198] Suppose that denotes The -point closed
Newton-Cotes formula with . There exist for which

(2.7)
If is even and and

(2.8)
If is odd and
Using this theorem with (2.8) returns

From (2.3) with , we obtain the standard trapezoidal rule

with its error term,

(2.9)
where .
Similarly, we obtain the standard Simpson’s 1/3 rule with its error term using (2.4) with
and (2.7) becomes,
 (2.10)
where .

Simpson’s 3/8 rule with its error term is obtained using (2.5) with
and (2.8) yields,

(2.11)
where
These indicates that the degree of precision of (2.7) is and (2.8) is although the
interpolation polynomial is of degree at most .

2.4. Composite Rules

Composite quadrature rules for trapezoidal and Simpson’s are simpler approach and is
considered an effective way to increase the accuracy of the result when integrating high-degree
polynomials. These rules involve partitioning the interval into subintervals then using the
low-order interpolants such as the standard Newton-Cotes methods on each subinterval.

2.4.1. Composite Trapezoidal Rule

Suppose interval is partitioned into intervals [ by
. Employing the standard trapezoidal rule on subinterval
returns

(2.12)

Since application of standard Newton-Cotes methods consider that is evaluated at equally

spaced nodes between and , then
(2.12) returns the composite trapezoidal rule,

(2.13)
Theorem2.3 [3, p. 206]Let , and .
There exists for which the Composite Trapezoidal rule for subintervals can be
written with its error term as

(2.14)

The error term in (2.14) can be derived from the summation of all errors in the application of
standard trapezoidal rule on each subinterval. Thus,

for .

Since then by Extreme Value Theorem,

we have, n

and,

By Intermediate Value Theorem, there is such that

Hence,

and since , then

It is intriguing to note that the error term for the composite trapezoidal rule is not
that we have in (2.9). These are not comparable because for standard trapezoidal rule is fixed
at since , but for composite trapezoidal rule for positive integer .

2.4.2. Composite Simpson’s Rule

Composite Simpson’s 1/3 rule is obtained in an equivalent manner (as Composite Trapezoidal
rule); however, we choose an even integer , partitioning the interval into
subintervals, then applying the standard Simpson’s rule on each consecutive pair of
subintervals. Applying some algebraic manipulations, we obtain

(2.15)

Theorem 2.4 [3, p. 206] Let be even , and

. There exists for which the composite
Simpson’s rule for subintervals can be written with its error term as

(2.16)
Similarly, the error term in (2.16) can be derived from the summation of all errors in the
application of standard Simpson’s rule on each consecutive pair of subintervals. Thus,

for for .

Since then by Extreme Value Theorem,

we have,

and

By Intermediate Value Theorem, there is such that,

Hence

and since , then

We can also observe that the error term for the Composite Simpson’s rule is , instead of
which was the error in the standard Simpson’s rule (2.10). This is simply because the
in standard Simpson’s rule is fixed at rule while in composite Simpson’s rule it is rule
for an even integer

2.5. MATHEMATICA Software

MATHEMATICA is considered a definitive system for modern technical computing [18]. It is
widely used in highly computational environments (both numerical and symbolic) because of its
robust and comprehensive features, which allow flexibility and reliability. We use
MATHEMATICA not only to ensure accuracy and precision of the results but also to visualize
and observe the behaviours of the functions in the three methods employed.

3. RESULTS AND DISCUSSION

This study investigates the robustness and accuracy of composite trapezoidal rule, Simpson’s
1/3 rule, and Simpson’s 3/8 rule to approximate the definite integrals of the standard normal
probability density function, incomplete gamma function, the Fresnel sine function, and
Dawson’s integral These functions hold significant importance and find extensive applications
in various disciplines such as Statistics, Business, and Engineering. The outcomes are discussed
in this section.

3.1 Standard Normal Probability Density Function

The standard normal probability density function is given by . The function is

the basis of calculating the area under the normal curve by determining the definite integral
given a particular interval. However, due to the complexity of the function, it is tedious to
determine the integral. For this reason, numerical integration is used to estimate the value within
a given domain.
Table 1 displays the results of approximating standard normal probability density function from
0 to 1. This is particularly the area of the normal curve from to
The table shows that both Simpson’s 1/3 rule and 3/8 rules are the most
efficient in terms approximating the integral since it only takes 2 subintervals, and
for the 1/3 rule and and for the 3/8 rule for the approximate value to
converge with actual value considering the absolute errors within tolerance level. The
trapezoidal rule on the other hand requires a significantly larger number of subintervals to
achieve a comparable level of accuracy, requiring 21 subintervals to approach the exact value.
 Table 1. Approximation of within

Trapezoidal Rule Simpson's 1/3 Rule Simpson's 3/8 Rule

n Approx. Exact Abs. Approx. Exact Abs. Approx. Exact Abs.
Answer Answer Error Answer Answer Error Answer Answer Error
0.320456 0.341344
1 5025 7461 0.0209
0.336260 0.341344 0.341529 0.341344
2 9146 7461 0.0051 0520 7461 0.0002
0.339095 0.341344 0.341425 0.341344 0.000
3 9119 7461 0.0022 8380 7461 1
0.340081 0.341344 0.341355 0.341344
4 8445 7461 0.0013 4879 7461 0.0000
0.340537 0.341344
5 0985 7461 0.0008
0.340784 0.341344 0.341349 0.341344 0.000
6 1090 7461 0.0006 5080 7461 0
0.340932 0.341344
7 9509 7461 0.0004
0.341029 0.341344
8 5157 7461 0.0003
0.341095 0.341344
9 7025 7461 0.0002
0.341143 0.341344
10 0365 7461 0.0002
0.341178 0.341344
11 0536 7461 0.0002
0.341204 0.341344
12 6843 7461 0.0001
: : :
-do-
: : :
0.341299 0.341344
21 0187 7461 0.0000

Although both Simpson’s 1/3 and 3/8 rules require the same number of subintervals to achieve
convergence, a closer examination reveals that the Simpson’s 3/8 rule yields a smaller absolute
error before reaching convergence with the true value. It can be noted in Figure. 1 that there is a
significant difference of the absolute error of the trapezoidal Rule compared to the other
methods. A large gap of the errors is observed on the initial subintervals until the desired
tolerance level is attained. On the other hand, there is a marginal difference between the errors
of 1/3 rule and 3/8 rule.
 Figure 1. Absolute Error Graphs

Below are the Mathematica outputs of approximation of within :

Absolute error set to be below:1.*10^-10
Searching for n that satisfies absolute error less than:1.*10^-10
**********************TRAPEZOIDAL RULE****************************
TOL met at n=9000, Absolute error at:2.489414785*10^-10
n=9000 approx: 0.3413447458 exact: 0.3413447461 Abs error : 2.489414785*10^-10
**********************SIMPSONS 1/3 RULE****************************
TOL met at n=74, Absolute error at:8.966438703*10^-11
n=74 approx: 0.3413447458 exact: 0.3413447461 Abs error : 8.966438703*10^-11
**********************SIMPSONS 3/8 RULE****************************
TOL met at n=90, Absolute error at:9.220813002*10^-11
n=90, approx: 0.3413447458 exact: 0.3413447461 Abs error : 9.220813002*10^-11
As revealed by the outputs, if we increase the tolerance level for absolute error to be within
, there is a significance increase of the values of the subintervals across the three methods
for the approximate value to converge with the exact value. Specifically, it necessitates 9000
subintervals for the trapezoidal rule and 37 and 30 for 1/3 and 3/8 Simpsons rules, respectively.
In the field of numerical integration, higher tolerance level indicates better precision and
accuracy.

3.2. Fresnel Sine Function

The Fresnel sine integral, is widely used in the field of optics and
used in the calculation of electromagnetic field density. The approximation of the function at
is scrutinized in this study. The results in Table 2 indicates that there is a lesser variation
in terms of the convergence behaviour across the 3 methods in approximating the definite
integral considering the tolerance level of . It can be observed from the results that it only
requires fewer subintervals of the intervals to achieve an approximation that closely aligns with
the true value. The trapezoidal rule necessitated the use of 6 subintervals to produce an
approximate value that converged with the exact value. In contrast, both Simpson’s 1/3 and 3/8
rules only required 4 to reach a similar level of accuracy.
 Table 2. Approximation of within

Trapezoidal Rule Simpson's 1/3 Rule Simpson's 3/8 Rule

n Approx. Exact Approx. Exact Approx. Exact
Error Error Error
Answer Answer Answer Answer Answer Answer
0.50000 0.43825 0.06
1 00000 91474 17
0.44134 0.43825 0.00 0.42178 0.43825 0.01
2 17162 91474 31 8955 91474 65
0.43881 0.43825 0.00 0.43116 0.43825 0.00
3 19291 91474 06 34203 91474 71
0.43842 0.43825 0.00 0.43745 0.43825 0.00
4 77565 91474 02 6437 91474 08
0.43832 0.43825 0.00
5 70254 91474 01
0.43829 0.43825 0.00 0.43811 0.43825 0.00 0.43791 0.43825 0.00
6 15734 91474 00 8121 91474 01 03055 91474 03
0.43821 0.43825 0.00
8
6495 91474 00
0.43819 0.43825 0.00
9
71694 91474 01
0.43824 0.43825 0.00
12
03145 91474 00

Additionally, the comparative analysis of the absolute errors across the three methods displayed
in Figure 2 reveals that the Simpson's 3/8 rule consistently yields the minimal magnitudes of
absolute errors until the desired precision is achieved. The first subinterval signifies larger
disproportions on the absolute errors. The gaps are minimized when the succeeding subintervals
are considered.

Figure 2. Absolute Error Graphs

Below are the Mathematica outputs of approximation of within when
the tolerance level is increased to . In contrast to the results using the marginal
difference of the subintervals is maintained across the three methods. As per the table, the
number of subintervals needed to maintain a comparable level of accuracy are: 142 for
trapezoidal rule, 101 for Simpson’s 1/3 rule, and 83 for 3/8 rule.
Absolute error set to be below:1.*10^-10
Searching for n that satisfies absolute error less than:1.*10^-10
**********************TRAPEZOIDAL RULE****************************
TOL met at n=142, Absolute error at:9.834355552*10^-11
n=142 approx: 0.4382591475 exact: 0.4382591474 Abs error : 9.834355552*10^-1
**********************SIMPSONS 1/3 RULE****************************
TOL met at n=202, Absolute error at:9.880962715*10^-11
n=202 approx: 0.4382591475 exact: 0.4382591474 Abs error : 9.880962715*10^-11
**********************SIMPSONS 3/8 RULE****************************
TOL met at n=249, Absolute error at:9.6294972*10^-11
n=249, approx: 0.4382591475 exact: 0.4382591474 Abs error : 9.6294972*10^-11

3.3. Incomplete Gamma Function

The incomplete gamma function given by is commonly applied in
the field of heat conduction, probability theory, and Fourier and Laplace transformations. The
convergence analysis to approximate the value of the incomplete gamma function utilizing
on the interval within is presented in Table 3.

Table 3. Approximation of within

Trapezoidal Rule Simpson's 1/3 Rule Simpson's 3/8 Rule

n Approx. Exact Approx. Exact Approx. Exact
Error Error Error
Answer Answer Answer Answer Answer Answer
0.18393 0.26424 0.08
1 97206 11177 03
0.24360 0.26424 0.02 0.26349 0.26424 0.00
2 25252 11177 06 01268 11177 08
0.25502 0.26424 0.00 0.26390 0.26424 0.00
3 05234 11177 92 56237 11177 03
0.25904 0.26424 0.00 0.26419 0.26424 0.00
4 50402 11177 52 25452 11177 00
0.26091 0.26424 0.00
5 28085 11177 33
0.26192 0.26424 0.00 0.26421 0.26424 0.00
6 87269 11177 23 95021 11177 00
0.26254 0.26424 0.00
7 17462 11177 17
0.26293 0.26424 0.00
8 98017 11177 13
0.26321 0.26424 0.00
9 27902 11177 10
0.26340 0.26424 0.00
10 80987 11177 08
 Trapezoidal Rule
n Approx. Exact
Error
Answer Answer
0.26355 0.26424 0.00
11 26272 11177 07
0.26366 0.26424 0.00
12 25656 11177 06
0.26374 0.26424 0.00
13 81311 11177 05
0.26381 0.26424 0.00
14 60294 11177 04
0.26387 0.26424 0.00
15 08094 11177 04
0.26391 0.26424 0.00
16 56448 11177 03
0.26395 0.26424 0.00
17 28047 11177 03
0.26398 0.26424 0.00
18 39460 11177 03
0.26401 0.26424 0.00
19 03015 11177 02
: : :
-do-
: : :
0.26409 0.26424 0.00
24 64512 11177 01
: : :
-do-
: : :
0.26419 0.26424 0.00
41 15451 11177 00

There is an exceptionally large discrepancy in terms of the robustness of the trapezoidal rule as
compared to Simpson’s 1/8 and 3/8 rules. To arrive at the desired accuracy level, the trapezoidal
rule required more than 20 times the number of subintervals of the Simpsons rules, and
subintervals for the 1/3 rule and and for the 3/8 rule.

Figure 3. Absolute Error Graphs

Relative to the results of the other two functions, the 3/8 rule maintained the lowest absolute
error as the value converges the exact answer as displayed in Figure 3. It can be noted from the
graph that in the first few values of subintervals, the absolute errors of utilizing the trapezoidal
method are significantly larger in contrast to the other 2 methods. Simpson’s 1/3 and 3/8 rules
exhibits no significant difference in the errors.

Mathematica outputs of approximation of within below compares the

trapezoidal rule, Simpson's 1/3 rule, and Simpson's 3/8 rule highlighting the number of
subintervals needed to achieve an absolute error when the tolerance level is within The
trapezoidal rule required 9000 subintervals to converge with the true value indicating slow
convergence. This is a significant increase when compared to the tolerance level.
Simpson's 1/3 rule achieved the target accuracy with 52 subintervals while the Simpson's 3/8
rule met the tolerance with 44 subintervals.
Absolute error set to be below:1.*10^-10
Searching for n that satisfies absolute error less than:1.*10^-10
**********************TRAPEZOIDAL RULE****************************
TOL met at n=9000, Absolute error at:1.028806873*10^-9
n=9000 approx: 0.2642411166 exact: 0.2642411177 Abs error : 1.028806873*10^-9
**********************SIMPSONS 1/3 RULE****************************
TOL met at n=106, Absolute error at:9.963702086*10^-11
n=106 approx: 0.2642411166 exact: 0.2642411177 Abs error : 9.963702086*10^-11
**********************SIMPSONS 3/8 RULE****************************
TOL met at n=132, Absolute error at:9.322431715*10^-11
n=132, approx: 0.2642411166 exact: 0.2642411177 Abs error : 9.322431715*10^-11

3.4. Dawson’s Integral

The Dawson’s integral, , bears important contribution in the concept
of heat conduction and the theory of electrical oscillations of special vacuum tubes. Table 4
shows the results when approximating the value of the Dawson’s integral on the interval
Comparative analysis with the results obtained in incomplete gamma function shows the same
convergence behaviour using the trapezoidal rule. The data further reveals that it takes 58
subintervals to attain the targeted level of precision. The magnitude is significantly larger as
compared to the outcomes of Simpson’s 1/3 and 3/8 rules. It is apparent on the results that the
3/8 rule is the most efficient method, requiring only 3 subintervals ( , 6, and 9), in
comparison to the 1/3 rule, which demands 4 subintervals ( , 4, 6, and 8) for the
approximate answer to converge with the exact answer.

Table 4. Approximation of within

Trapezoidal Rule Simpson's 1/3 Rule Simpson's 3/8 Rule

n Approx. Exact Approx. Exact Approx. Exact
Error Error Error
Answer Answer Answer Answer Answer Answer
0.683939 0.538079
1 0.1459
7206 5069
0.578153 0.538079 0.542890 0.53807
2 0.0401 0.0048
1367 5069 9420 95069
0.556268 0.538079 0.54030 0.53807
3 0.0182
4773 5069 95719 95069 0.0022
 0.548390 0.538079 0.538469 0.53807
4 0.0103 0.0004
1066 5069 0966 95069
0.544702 0.538079
5 0.0066
3760 5069
0.542687 0.538079 0.538161 0.53807 0.53825 0.53807
6 0.0046 0.0001
9211 5069 0690 95069 47691 95069 0.0002
0.541469 0.538079
7 0.0034
3850 5069
0.540676 0.538079 0.538105 0.53807
8 0.0026 0.0000
9309 5069 8724 95069
0.540132 0.538079 0.53811 0.53807
9 0.0021
9056 5069 59591 95069 0.0000
0.539743 0.538079
10 0.0017
4061 5069
0.539455 0.538079
11 0.0014
0259 5069
0.539235 0.538079
12 0.0012
5782 5069
0.539064 0.538079
13 0.0010
7298 5069
0.538929 0.538079
14 0.0008
1253 5069
0.538819 0.538079
15 0.0007
6999 5069
0.538730 0.538079
16 0.0007
1253 5069
0.538655 0.538079
17 0.0006
8760 5069
0.538593 0.538079
18 0.0005
6459 5069
0.538540 0.538079
19 0.0005
9745 5069
0.538496 0.538079
20 0.0004
0001 5069
0.538457 0.538079
21 0.0004
2932 5069
0.538423 0.538079
22 0.0003
7410 5069
: : :
-do-
: : :
0.538325 0.538079
26 0.0002
9945 5069
: : :
-do-
: : :
0.538223 0.538079
34 0.0001
6615 5069
: : :
-do-
: : :
0.538129 0.538079
58 0.0000
0487 5069
As observed from Figure 4, the absolute errors of the trapezoidal rule considering the first few
intervals exhibits a wide gap as compared to Simpson’s 1/3 and 3/8 rules. Although there is a
marginal difference on the errors between 1/3 and 3/8 rules, the latter bagged the lowest
magnitude.

Figure 4. Absolute Error Graphs

When the tolerance level is increased to it can be observed from Table 8 that trapezoidal
rule requires 9000 subintervals to achieve convergence. This is comparable to the outcomes of
the incomplete gamma function and standard normal distribution function. Simpson’s 3/8 Rule
has the fastest convergence which necessitates the use of 75 subintervals while the 1/3 Rule
requires 92 subintervals.
Absolute error set to be below:1.*10^-10
Searching for n that satisfies absolute error less than:1.*10^-10
**********************TRAPEZOIDAL RULE****************************
TOL at n=9000, Absolute error at:2.057612525*10^-9
n=9000 approx: 0.538079509 exact: 0.5380795069 Abs error : 2.057612525*10^-9
**********************SIMPSONS 1/3 RULE****************************
TOL met at n=184, Absolute error at:9.693024161*10^-11
n=184 approx: 0.538079509 exact: 0.5380795069 Abs error : 9.693024161*10^-11
**********************SIMPSONS 3/8 RULE****************************
TOL met at n=225, Absolute error at:9.753819974*10^-11
n=225, approx: 0.538079509 exact: 0.5380795069 Abs error: 9.753819974*10^-11

4. CONCLUSIONS
The results in the preceding section reveal distinct patterns regarding the efficiency and
accuracy of various numerical integration methods across different classes of functions.
Simpson’s 1/3 and 3/8 rules consistently surpass the trapezoidal rule. The two methods show
faster convergence and higher level of accuracy. However, when the tolerance level is increased
to a considerable magnitude, the Simpson’s 3/8 rule emerged as the most robust among the
three methods. Moreover, comparative analysis of absolute errors reveals that the Simpson’s 3/8
rule yields smaller magnitudes of absolute errors. Conversely, the trapezoidal rule consistently
exhibits a slower convergence rate, which requires larger number of subintervals to attain the
desired level of accuracy.
The four functions utilized a common interval for estimating the definite integral, thus it
is recommended to investigate various domains to substantiate the findings and to gain a deeper
understanding on the numerical methods utilized in the study. In addition to examining absolute
error, future research should incorporate a comprehensive error analysis that includes relative
error, truncation error, and round-off error and error bound. This can provide a more detailed
understanding of the sources and magnitudes of errors in numerical integration. Furthermore,
we recommend extending the research to include higher-dimensional integrals could provide
valuable insights into the robustness of these numerical methods.
ACKNOWLEDGEMENTS
The authors would like to express their deepest gratitude to Dr Rito Opol Jr. for his valuable
suggestions during the course this work. Thank you, Dr Rito, for your unwavering support.

REFERENCES
[1] D. Huybrechs, “Stable high-order quadrature rules with equidistant points,” Journal of
Computational and Applied Mathematics, vol. 231, no. 2, pp. 933–947, Sep. 2009, doi:
10.1016/j.cam.2009.05.018.
[2] “Newton-Cotes Formulas -- from Wolfram MathWorld.”
https://mathworld.wolfram.com/Newton-CotesFormulas.html
[3] R. L. Burden and D. Faires, Numerical Analysis, 9th ed. USA: BROOKS/COLE CENGAGE
Learning, 2011.
[4] “Numerical Quadrature,” Google Books.
https://books.google.com.bh/books/about/Numerical_Quadrature.html?id=yWaE0AEACAAJ&r
edir_esc=y
[5] M. Embree, “Lecture Notes on Numerical Analysis.” https://personal.math.vt.edu/embree
(accessed Jun. 10, 2024).
[6] Md. J. Uddin, N. M. Md. Moheuddin, and N. Md. Kowsher, “A New Study of Trapezoidal,
Simpson’s 1/3 and Simpson’s 3/8 Rules of Numerical Integral Problems,” Applied Mathematics
and Sciences: An International Journal, vol. 6, no. 4, pp. 01–13, Dec. 2019, doi:
10.5121/mathsj.2019.6401. https://doi.org/10.5121/mathsj.2019.6401
[7] “Newton–Cotes Formulas for Numerical Integration in Maple - International Journal of
Mathematics and Statistics Studies (IJMSS),” International Journal of Mathematics and Statistics
Studies (IJMSS), Feb. 28, 2023. https://eajournals.org/ijmss/vol10-issue-2-2022/newton-cotes-
formulas-for-numerical-integration-in-maple/
[8] A. H. Nzokem, “Numerical solution of a Gamma - integral equation using a higher order
composite Newton-Cotes formulas,” Journal of Physics. Conference Series, vol. 2084, no. 1, p.
012019, Nov. 2021, doi: 10.1088/1742-6596/2084/1/012019.
[9] S. S. Olver, “Numerical approximation of highly oscillatory integrals,” Jul. 08, 2008.
https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.612289
[10] H. M. Golshan, “Numerical solution of nonlinear m-dimensional Fredholm integral equations
using iterative Newton–Cotes rules,” Journal of Computational and Applied Mathematics, vol.
448, p. 115917, Oct. 2024, doi: 10.1016/j.cam.2024.115917.
[11] P. Magalhaes and C. Magalhaes, “Higher-Order Newton-Cotes Formulas,” Journal of
Mathematics and Statistics, vol. 6, no. 2, pp. 193–204, Apr. 2010, doi:
10.3844/jmssp.2010.193.204.
[12] A. J. Ali and A. F. Abbas, “Applications of Numerical Integrations on the Trapezoidal and
Simpson’s Methods to Analytical and MATLAB Solutions,” Mathematical Modelling and
Engineering Problems/Mathematical Modelling of Engineering Problems, vol. 9, no. 5, pp.
1352–1358, Dec. 2022, doi: 10.18280/mmep.090525.
[13] J. Roumeliotis, “Numerical Integration - A Maple Perspective,” Jan. 2003, [Online]. Available:
https://vuir.vu.edu.au/17848/
[14] E. Sermutlu, “Comparison of Newton–Cotes and Gaussian methods of quadrature,” Applied
Mathematics and Computation, vol. 171, no. 2, pp. 1048–1057, Dec. 2005, doi:
10.1016/j.amc.2005.01.102.
[15] J. Wu and W. Sun, “The superconvergence of Newton–Cotes rules for the Hadamard finite-part
integral on an interval,” Numerische Mathematik, vol. 109, no. 1, pp. 143–165, Dec. 2007, doi:
10.1007/s00211-007-0125-7.
[16] C. O. E. Burg, “Derivative-based closed Newton–Cotes numerical quadrature,” Applied
Mathematics and Computation, vol. 218, no. 13, pp. 7052–7065, Mar. 2012, doi:
10.1016/j.amc.2011.12.060.
[17] J. C. Aguilar, “Higher-order Newton–Cotes rules with end corrections,” Applied Numerical
Mathematics, vol. 88, pp. 66–77, Feb. 2015, doi: 10.1016/j.apnum.2014.10.004.
[18] “WOLFRAM MATHEMATICA.” https://www.wolfram.com/mathematica/?source=nav
(accessed Jun. 04, 2024).

Authors
Lea A. Catapang is a seasoned mathematics
educator with a Master of Science in
Mathematics degree. She is currently teaching
mathematics and IT courses at Bahrain
Polytechnic's School of Foundation. She has
dedicated fourteen years to university-level
education and twelve years to a polytechnic
setting. Her experience in both academic
environments made her a more versatile and
effective educator.

James Ray Mendaje is an accomplished

mathematics educator with almost two decades
of extensive experience in both university and
polytechnic environments. He holds a Master of
Arts degree in Education with a major in
Mathematics and is currently teaching at the
School of Foundation, Bahrain Polytechnic. His
experience working as a quality assurance
specialist has significantly contributed to
academic excellence and student success.

Venus D. Siong is a committed educator with

over two decades of specialised knowledge in
mathematics. She proved her unwavering
dedication to the advancement of education by
earning her Doctor of Philosophy in
Mathematics Education. She is currently an
associate professor at Don Mariano Marcos
Memorial State University, where she teaches
graduate and undergraduate courses and
oversees instructional materials and
development.