1

Applications of Calculus III Ideas to Probability

Makenna Clark, Ellery Mays, and Sarah Steflik

Dr. Giambrone

MAT 3010 Calculus III

7 December 2023
 2

Definitions and Background

Calculus II courses typically cover how integrals can be used to study probability. In this

document, we will explore how the idea of normal distributions in probability and statistics

connects to integrals, series, and power series from calculus. Below are some key definitions and

examples that will be referred to.

Definition 1: A continuous random variable, called X, is a variable whose value can range

throughout real numbers but may only be measured or recorded to the nearest integer. We often

want to compute the probability that our continuous random variable is in a given range of

values.

Notation 1: The probability that the continuous random variable X is in the interval [a, b] is

denoted by 𝑃(𝑎 ≤ 𝑋 ≤ 𝑏).

Example 1: Let X be the cholesterol level of a randomly chosen person between the ages of 10

and 19. The cholesterol level comes from a range of real numbers but is measured to the nearest

whole number. For children between the ages of 10 and 19, cholesterol levels between good and

high (called “borderline”) are levels in the range 170 to 199. This means that

𝑃(170 ≤ 𝑋 ≤ 199) the probability that a random person's cholesterol level between the ages of

10 and 19 is borderline.
 3

Definition 2: Let f(x) be a continuous function that is defined for all −∞ < x < ∞. Using

properties of integrals and the definition of improper integral we’ve used in class, we get the

following definition for a new type of improper integral.

∞ 𝑏
∫ 𝑓(𝑥)𝑑𝑥 = lim lim ∫ 𝑓(𝑥)𝑑𝑥.
−∞ 𝑎 → −∞ 𝑏 → ∞ 𝑎

Definition 3: Every continuous random variable X has a probability density function f(x) that

must satisfy the following three properties.

𝑏
1. 𝑃(𝑎 ≤ 𝑋 ≤ 𝑏) = ∫ 𝑓(𝑥)𝑑𝑥.
𝑎

2. f(x) ≥ 0 for all x.

∞
3. P(−∞ < X < ∞) = ∫ 𝑓(𝑥)𝑑𝑥 = 1.
−∞

Property 1 above explains how definite integrals can be used to compute probabilities. Property 2

above says that a probability density function cannot be negative. Property 3 above explains the

intuitive idea that there is a 100% chance (a probability of 1) that our measurement of X is in the

range of all real numbers. This is like saying that the probability of getting 1, 2, 3, 4, 5, or 6

when you roll a die is 1 since there is a 100% chance that the outcome will be one of the values

1, 2, 3, 4, 5, or 6. We now use integrals to define the mean of a probability density function.

Definition 4: The mean of a probability density function f(x) is defined to be the improper

∞
integral ∫ 𝑥 · 𝑓(𝑥)𝑑𝑥. One of the most important probability density functions is associated
−∞

with the normal distribution, which models many random phenomena in real life.
 4

Definition 5: The probability density function for the normal distribution of a continuous
2 2
1 −(𝑥−µ) /(2σ )
random variable X is 𝑔(𝑥) = 𝑒 . The letter µ is used for the mean of the
σ 2Π

probability density function and the letter σ is used for the standard deviation of the probability

density function. Roughly speaking, µ represents the “expected value” of the random variable X,

and σ measures how “spread out” the values of the random variable X are.

2 2
1 −(𝑥−µ) /(2σ )
Notation 2: Given the probability density function 𝑔(𝑥) = 𝑒 for the normal
σ 2π

2
1 −𝑥 /2
distribution, let 𝑓(𝑥) = 𝑒 denote a special version of g(x) where µ = 0 and σ = 1.
2Π

Part 1: Integrals and Probability

Part 1 of this document is about applying definite integrals, improper integrals, and area

ideas to the study of probability.

2
∞ 𝑥
1 −
Using the definitions above, we can compute the improper integral ∫ 𝑥 · 𝑒 2
𝑑𝑥.
2π
−∞

2
∞ 𝑥
1 −
The indefinite integral ∫ 𝑥 · ·𝑒 2
𝑑𝑥 can be rewritten using Definition 2 as below.
2π
−∞

𝑏
1 1
lim lim ∫ 𝑥 · · 𝑑𝑥
2π 2
𝑎 → −∞ 𝑏 → ∞ 𝑎 𝑒
𝑥
 5

Using algebra to simplify, the now definite integral can be rewritten as an integral evaluated with

limits to infinity.

𝑏
1 𝑥
lim lim ∫ · 𝑑𝑥 .
2π 2
𝑎 → −∞ 𝑏 → ∞ 𝑎 𝑒
𝑥

By multiplying by 1, or 2 over 2, the integral can now be rewritten using u-substitution. These

next steps show the substitution process.

2
µ =𝑥 𝑑µ = 2𝑥𝑑𝑥

𝑏
1 1 2𝑥
2
· lim lim ∫ · 𝑑𝑥
2π 2
𝑎 → −∞ 𝑏 → ∞ 𝑎 𝑒
𝑥

𝑏
1 2𝑥
= · lim lim ∫ 𝑑𝑥
2 2π 2
𝑎 → −∞ 𝑏 → ∞ 𝑎 𝑒
𝑥

𝑥=𝑏
1 1
= · lim lim ∫ 𝑑µ
2 2π µ
𝑎 → −∞ 𝑏 → ∞ 𝑥=𝑎 𝑒

Using algebraic techniques, the integral is simplified to be computable.

𝑥=𝑏
1 1
= · lim lim ∫ µ 𝑑µ
2 2π
𝑎 → −∞ 𝑏 → ∞ 𝑥=𝑎 𝑒2

𝑥=𝑏 µ
1
= · lim lim ∫ − 𝑒 2 𝑑µ
2 2π
𝑎 → −∞ 𝑏 → ∞ 𝑥=𝑎

Now computing the integral, the expression can be shown in terms of a and b as shown below.

( )
µ
1 2 𝑥=𝑏
= · lim lim µ
𝑒 2 |𝑥=𝑎
2 2π
𝑎 → −∞ 𝑏 → ∞

( )
2
𝑥
1 2 𝑏
= · lim lim 2 𝑒 2
|𝑎
2 2π 𝑥
𝑎 → −∞ 𝑏 → ∞

( )
2 2
𝑏 𝑎
1 2 2
= · lim lim 2 𝑒 2
− 𝑎
𝑒 2

2 2π 𝑏
𝑎 → −∞ 𝑏 → ∞
 6

Using the solved integral above, the limits can now be solved as below.

( ( ) )
2 2
𝑏 𝑎

= 1
2 2π
· lim lim
2
2 𝑒 2
−
2
𝑎
𝑒 2

𝑎 → −∞ 𝑏 → ∞ 𝑏

(( ) )
2 2
∞ 𝑎
1 2 2
= · lim 2 𝑒 2
− 𝑎
𝑒 2

2 2π ∞
𝑎 → −∞

(( ) )
2
𝑎
1 ∞ 2
= · lim 0 ·𝑒 − 𝑎
𝑒 2

2 2π
𝑎 → −∞

( )
2
𝑎
1 2
= · lim 0 − 𝑎
𝑒 2

2 2π
𝑎 → −∞

( )
2
∞
1 2
= · 0− ∞
𝑒 2

2 2π

∞
=
1
2 2π
(
· 0 − 0 ·𝑒 )
1
= · (0 − 0)
2 2π

=0

This computation demonstrates that the mean of the probability density function
2
𝑥
1 −
𝑓(𝑥) = 𝑒 2
is 0. We can see in Definition 4 that the mean of f(x) is defined as
2π

2 2
∞ 𝑥 ∞ 𝑥
− −
∫ 𝑥 · 𝑓(𝑥) 𝑑𝑥. So, 𝑓(𝑥) =
1
2π
𝑒 2
, the mean would be ∫ 𝑥 · 1
2π
𝑒 2
𝑑𝑥.
−∞ −∞

Because the indefinite integral was found above to be equal to 0, the mean of the
2
𝑥
1 −
probability density function 𝑓(𝑥) = 𝑒 2
is also equal to zero.
2π
 7

Below is a graph of the function 𝑓(𝑥).

2
1 𝑥
1 −
Integrating our probability density function tells us that ∫ 𝑒 2
𝑑𝑥 =
2π
0

0. 341344746069, meaning there is a 34% probability that X is in the interval [0,1].

Using the idea that the definite integral of a nonnegative function is the area below the

curve over the region of integration, we can say the probability that X is in the interval [0,1]

represents the area under the curve between x=0 and x=1. The probability is the idea that the area

between x=0 and x=1 is 34% of the entire area under the curve.
 8

Below is a graph of the function 𝑓(𝑥) with the shaded region representing the area under the

curve from [0,1].

The integrals below show the probability that X is in the interval [3,∞).

∞ 𝑏
∫ 𝑓(𝑥)𝑑𝑥 = lim ∫ 𝑓(𝑥)𝑑𝑥
3 𝑏→∞ 3

2 2
∞ 𝑥 𝑏 𝑥
1 − 1 −
∫ 𝑒 2
𝑑𝑥 = lim ∫ 𝑒 2
𝑑𝑥
2π 2π
3 𝑏→∞ 3

Using the idea that the definite integral of a nonnegative function is the area below the

curve over the region of integration, we can say the probability that X is in the interval [0,3]

represents the area under the curve between x=0 and x=3. The probability is the idea that the area

between x=0 and x=3 is the probability percentage of the entire area under the curve.
 9

Below is a graph of the function 𝑓(𝑥) with the non-shaded region representing the area under the

curve from [3,∞].

The following integrals show the probability that X is in the interval (−∞,∞).

∞ 𝑏
∫ 𝑓(𝑥)𝑑𝑥 = lim lim ∫ 𝑓(𝑥)𝑑𝑥
−∞ 𝑎 → −∞ 𝑏 → ∞ 𝑎

Using the idea that the definite integral of a nonnegative function is the area below the

curve over the region of integration, we can say the probability that X is in the interval [-∞,∞]

represents the total area under the curve. The probability is the idea that the area is the

probability percentage of the entire area under the curve.

∞
Using definition 3, we know that the area under the curve represented by ∫ 𝑓(𝑥)𝑑𝑥 is 1.
−∞
 10

Part 2: Power Series, Infinite Series, and Probability

In this part, we will apply our knowledge from Calculus III to the normal distribution and its
2
1 −𝑥 /2
probability density function f(x) = 𝑒 . In particular, we will use what we have learned to:
2π

● represent indefinite integrals by power series,

● represent definite integrals by infinite series of real numbers,

● approximate sums S of infinite series by nth partial sums 𝑆𝑛 , and

● use an error approximation theorem to work with the error in using the nth partial sum 𝑆𝑛

to approximate the infinite sum S of an infinite series.

For the remainder of this document, we will assume all power series are anchored at a=0.

𝑥
Below is the power series for 𝑒 . The interval of convergence for this power series is −∞ < x < ∞.

∞ 𝑘
𝑥 𝑥
𝑒 = ∑ 𝑘!
𝑓𝑜𝑟 − ∞ < 𝑥 < ∞
𝑘=0

2
−𝑥 /2
Next, we will write out a power series 𝑒 .
2
−𝑥
Let u = 2

∞ 𝑘 2 3
𝑢 𝑢 𝑢 𝑢
𝑒 =∑ 𝑘!
= 1+ 𝑢 + 2!
+ 3!
...
𝑘=0

2 𝑘 2 2 3
2 ∞ −𝑥 2 −𝑥 2 (−𝑥 )
−𝑥 /2 ( 2
) 𝑥 ( 2
) 2
𝑒 =∑ 𝑘!
= 1 − 2
+ 2!
+ 3!
𝑘=0

2 𝑘
∞ 𝑥 2 4 6
𝑘 ( ) 𝑥 𝑥 𝑥
= ∑ (− 1) 2
𝑘!
=1 − 2
+ 8
− 48
𝑘=0
 11

2
Using the method of u-substitution with 𝑢 = − 𝑥 /2, we can attempt to compute the indefinite

2
−𝑥 /2 2
integral ∫ 𝑒 𝑑𝑥. However, this does not work because, for u = − 𝑥 /2, du would need to be

− 𝑥𝑑𝑥. This is impossible since there is no (-x) in the original function. Since using standard

integration methods to compute the indefinite integral does not work, we can use power series

instead. Using exponent rules, we can simplify our answers to get:

Short way:
2 𝑘
2 ∞ (𝑥 ) ∞ 𝑘 2𝑘+1
−𝑥 /2 𝑘 2 (−1) 𝑥
∫𝑒 𝑑𝑥 = ∫[ ∑ (− 1) 𝑘!
]= 𝐶 + ∑ 𝑘
𝑘=0 𝑘=0 2 (2𝑘+1)𝑘!

Long way:

2 2 4 6
−𝑥 /2 𝑥 𝑥 𝑥 1 3 1 5 1 7
∫𝑒 𝑑𝑥 = ∫[ 1 − 2
+ 8
− 48
]= 𝐶 + 𝑥 − 6
𝑥 + 40
𝑥 − 336
𝑥

Using the Ratio Test, we find the interval of convergence for our power series.

∞ 𝑘 2𝑘+1
(−1) 𝑥
∑ 𝑘
𝑘=0 2 (2𝑘+1)𝑘!

𝑘 2𝑘+1 𝑘+1 2𝑘+2

(−1) 𝑥 (−1) 𝑥
𝑎𝑘 = 𝑘 𝑎𝑘+1 = 𝑘+1
2 (2𝑘+1)𝑘! 2 (2𝑘+2)(𝑘+1)!

𝑘+1 2𝑘+2
| (−1) 𝑥 |
lim | 2
𝑘+1
(2𝑘+2)(𝑘+1)! |
| (−1) 𝑥
𝑘 2𝑘+1
|
𝑘→∞ | 𝑘
2 (2𝑘+1)𝑘! |

| (−1)𝑘+1𝑥2𝑘+2 (−1) 𝑥
𝑘 2𝑘+1
|
= lim | 𝑘+1 · |
𝑘→∞
| 2 (2𝑘+2)(𝑘+1)! 𝑘
2 (2𝑘+1)𝑘! |

| (−1) · 𝑥3 · (2𝑘+1) 𝑘! |
= lim | 2 · (2𝑘+2)(𝑘+1)! |
𝑘→∞
| |

| (−𝑥)3· (2𝑘+1)(𝑘)(𝑘−1)(𝑘−2)... |
= lim | 2(2𝑘+2)(𝑘+1)(𝑘)(𝑘−1)(𝑘−2)... |
𝑘→∞
| |
 12

| (−𝑥)3(2𝑘+1)2 |
= lim | 2(2𝑘+4+2) |
𝑘→∞
| |

3
= −𝑥| | lim
2𝑘+1
2
4𝑘 +8𝑘+4
𝑘→∞

2𝑘 1
3 +
| |
2 2
𝑘 𝑘
= −𝑥 lim 4𝑘
2
8𝑘 4
𝑘→∞ 𝑘
2 +
𝑘
2 + 2
𝑘

2 1
3 +
| |
𝑘 2

=−𝑥 lim 10
𝑘
6
4+ +
𝑘→∞ 𝑘 𝑘
2

3
= −𝑥 | |· 0
4

= 0

IOC = − ∞ < 𝑥 < ∞

Using our answer from above, we can write out an infinite series (of real numbers) for the

1 2
−𝑥 /2
definite integral ∫ 𝑒 𝑑𝑥. Using the original power series, we can write an infinite series of
0

1 2
−𝑥 /2
real numbers for the definite integral ∫ 𝑒 𝑑𝑥.
0

1
−𝑥/2
∫𝑒 dx
0

𝑘 2𝑘+1
(−1) 𝑥 𝑥=1
= 𝑘 |𝑥=0
2 (2𝑘+1)𝐾!

𝑘 2𝑘+1 𝑘 2𝑘+1
(−1) (1) (−1) (0)
=[ 𝑘 − 𝑘 ]
2 (2𝑘+1)𝑘! 2 (2𝑘+1)𝑘!

𝑘
(−1) · 1
= 𝑘
2 (2𝑘+1)𝑘!

𝑘 ∞
(−1) 𝑘 1 1 1 1
= 𝑘 → ∑ (− 1) · 𝑘 = 1− 3!
+ 20(2!)
− 56(3!)
+ ...
2 (2𝑘+1)𝑘! 𝑘=0 2 (2𝑘+1)𝑘!
 13

We are studying an alternating series.

1
We can use the alternating series test because the sequence { 𝑘 } is eventually
2 (2𝑘+1)𝑘!

decreasing as shown in the graph.

1 1
lim 𝑘 = ∞
= 0
𝑘→∞ 2 (2𝑘+1)𝑘!

1
Since the limit of 𝑘 is 0, and the sequence of the function is eventually decreasing, we
2 (2𝑘+1)𝑘!

∞
𝑘 1
can say that the series ∑ (− 1) · 𝑘 is convergent.
𝑘=0 2 (2𝑘+1)𝑘!

If we approximate S using the 6th partial sum 𝑆6 , we can find the associated error bound.

|
𝐸6 = 𝑆 − 𝑆6 ≤ 𝑎𝑛+1 |
7
𝑘
| |
𝐸6 = 𝑆 − 𝑆6 ≤ 𝑎7 → ∑ (− 1) ·
𝑘=0
𝑘
1
2 (2𝑘+1)𝑘!

−7
= −1
9676800
= − 1. 03339947 𝑋 10
 14

In general, getting an nth partial sum 𝑆𝑛 to agree with the infinite sum S to at least 8

−9
decimal places means we need the error to be less than 0.000000005 or 5 × 10 in scientific

notation.

1 2
1 −𝑥 /2
Let 𝑆𝑛 denote the nth partial sum of the series we found ∫ 𝑒 𝑑𝑥 previously. Using
2π
0

Desmos we can find the smallest possible value of n needed so that 𝑆𝑛 it correctly approximates

S to 8 decimal places.

f(1) Too big

f(5) Too big

f(6) Too big

f(8) Good

f(9) Too small

Next, we can use Desmos to compute 𝑆𝑛, where n is the value we found in part in the chart. Note

that our value 𝑆𝑛 approximates, to 8 decimal places, the probability that the continuous random

variable X is in the interval [0, 1] assuming we have the probability density function f(x)

associated with the normal distribution with mean µ = 0 and standard deviation σ = 1.

−9
𝑆8 = 5. 698894141 × 10 (Desmos)
 15

Conclusions

Based on what we have completed above, we can conclude that the definite integral
1 2
−𝑥 /2
∫𝑒 𝑑𝑥 cannot be computed exactly using integration techniques of Calculus II. Instead, we
0
1 2
−𝑥 /2
can approximate the value ∫ 𝑒 𝑑𝑥 by using the nth partial sum 𝑆𝑛 of an infinite series (of real
0
numbers), where n is the value we found in the chart above. As explored in Section 2, the
1 2
−𝑥 /2
definite integral ∫ 𝑒 𝑑𝑥 cannot be computed exactly using the integration techniques of
0
1 2
−𝑥 /2
Calculus II (u-substitution). Instead, we can approximate the value ∫ 𝑒 𝑑𝑥 by using the nth
0
partial sum 𝑆𝑛 of an infinite series (of real numbers). Furthermore, we know from our
1 2
−𝑥 /2
computations that our approximation 𝑆𝑛 agrees with the exact value of ∫ 𝑒 𝑑𝑥 to at least 8
0
decimal places. In general, we can choose n large enough so that the approximation error is as

small as we need. This is why the Alternating Series Estimation Theorem and the Lagrange Error

Bound Theorem is so important.

 16

References

Giambrone, A. (2023). Applications of Calculus III Ideas to Probability. MAT 3010 Calculus III.

https://elmira.instructure.com/courses/10459/assignments/120070

Desmos Links

https://www.desmos.com/calculator/sxpcifhe1h

https://www.desmos.com/calculator/nrpdqph5fw

https://www.desmos.com/calculator/c6crnrr48n