Continuous Random Variables

Table of Contents

Probability Models for Continuous Random Variables..........................................................3

PMF.....................................................................................................................................................3

CDF......................................................................................................................................................5

PDF......................................................................................................................................................6

Properties of PDFs........................................................................................................................11

Properties of CDFs........................................................................................................................11

Expectations and Variances.........................................................................................................15

Well-Known Continuous Random Variables...........................................................................16

Uniform Continuous Random Variables...............................................................................16

Exponential Random Variable.................................................................................................20

Gamma Random Variables.......................................................................................................24

Gaussian Random Variables....................................................................................................27

Beta Random Variables.............................................................................................................35

For a continuous random variable, the number of possible values is uncountable, and

thus is infinite. For example, if we were to select a random point on a line between 0

to 1, it could have infinite possible values, or when a person arrives, we can state the

time to hours, to minutes, to seconds, to milliseconds and so on.

If X is a continuous random variable, S X =( 0 , 1 ) (exclusive) or S X =[ 0 , 1 ] (inclusive).

Notice that unlike with discrete random variables, we are not giving a set of specific

values, since it is literally not possible. Essentially, we can be infinitely more specific.
Probability Models for Continuous Random Variables

PMF

For discrete random variables, we used PMF and CDF as probability models. However,

PMF is not applicable for continuous random variables. This is because P [ X=x ] =0 ,

always, for continuous random variables.

We can prove this. Consider, for simplicity, that all values for a continuous random
1
variable are equally likely. We know that P [ X=x ] = in this case. Since there are
n
infinite possible values, n=∞ , meaning P [ X=x ] is always 0 .

More formally, PMF is nothing more than distributing 1 unit of mass on a number line.

If we have an infinite number of points on the number line, then the allocation per

point is 0.

Consider a situation where we have a line AB of length 1. We will pick a random point
C on that line. Here, X =length of AC , so X is a continuous random variable and

S X =( 0 , 1 ). We know that P [ X=x ] .

Now consider the same situation, except that we have discretised the line, dividing it

into n equal parts. Thus, we can define a random variable

Y =number of the part where C is selected , and SY = {1 , 2 ,… n }. Due to this difference, Y is a
1
discrete random variable, and P [ Y = y ] = , given all the parts are equally likely to be
n
chosen.
There is a relationship between X and Y here. Say n=10 and we pick the point 0.62.
Y =⌈ nX ⌉=7 . In fact, for all values of X between 0.6 and 0.6 9 (notice that this is

recurring), Y =7 . The question is, how well does Y approximate X ? That depends on

how large the value of n is. The larger the value of n, the more accurately X can be

approximated by Y .

We can see that, for { X =x } (the PMF of X can also be represented like this

apparently) and {Y =⌈ nX ⌉ }, many values of X are represented by a single value of Y .

Thus, { X =x } ⊂ {Y =⌈ nX ⌉ }. As such,

P [ X=x ] ≤ P [ Y =⌈ nX ⌉ ]
1
P [ X=x ] ≤
n

Again, as we increase the value of n,

P [ X=x ] ≤ lim P [ Y =⌈ nX ⌉ ]
x→ ∞

1
≤ lim
n →∞ n
≤0

The very definition of the PMF probability model states that P [ X=x ] ≥ 0 . Thus, the

only takeaway we have from this is that for continuous random variables, P [ X=x ] =0 .

This was the mathematical proof of the same thing we discovered one whole page

and a headache ago.

I don’t know if you’ve noticed but, we’ve kind of been using PMF a lot. It going poof

here is a bit a problem for us. We stated that one of the goals of using random

variables was to facilitate further processing, and PMFs have been helping us a lot

with that. Now that they’re gone, we need to find something to replace them, and we

will do exactly that. Wait a bit though.

CDF

1
Say we pick the point 0.5 on the line of 1 meter. Here, P [ 0< X ≤ 0.5 ] = . This is
2
because, since the probability of picking any of the points is the same, we will be

picking one of the points to the left of 0.5 in half of the cases and one of the points to
1 1
the right in half of the cases. Similarly, P [ 0< X ≤ 0.25 ] = , P [−∞< X ≤ 0.5 ] = and P ¿.
4 2

Even if we cannot find the probability for specific values of X , we can find the

probability for an interval. P [ X ≤ x ] =F X ( x )=x . As such, CDF can be used with

continuous random variables.

{
0 x <0
F X ( x )= x 0< x <1
1 x≥1
A few points to notice:

 F X (−∞ )=P [ X ≤−∞ ]=0

 F X ( +∞ )=1
 P [ a< X ≤ b ]=F X ( b )−F X ( a )
 P [ X ≤b ] =P [ X <b ]

This last point is correct because P [ X ≤b ] =P [ X <b ] + P [ X=b ] and we know that
P [ X=b ] =0. Following the same logic:

 P [ a< X ≤ b ]=P [ a< X <b ] =P [ a ≤ X < b ]

PDF

Density is the mass per unit volume, but it is also possible to define it as a unit per

unit area, for example the density of people per unit area in a field. We can even find

examples for valid definitions of density per unit length.

The probability density function (PDF) defines the average probability per unit

length.

When we first started discussing random variables, we stated that one of its major

purposes was to allow further processing of the probabilities of the different

outcomes. One of the key ways we achieved this with discrete random variables was

by using PMFs. However, as we have seen, PMFs cannot be used with continuous

random variables. As such, we need a probability model that can work with

continuous random variables in the way that PMFs worked with discrete random

variables. This is where Probability Density Functions (PDFs) come in.

 {
0 x <0
Consider two cases of CDFs. In the first case, F X ( x )= x 0< x <1 and in the second
1 x≥1

{
0 x <0
x
case, F X ( x )= 0< x < 4 . If we draw the CDF graphs for both these cases, we will find
4
1 x≥4

that they both increase linearly, but the graph for the first case has a much steeper

slope.

We can understand this by considering the probability distributions as mass

distributions. Thus, in the first case, where x has values between 0 and 1, a total

probability of 1 is distributed over a smaller range, thus giving a steeper slope. In the

second case, the same total probability of 1 is distributed over a larger range,

between 0 and 4 , thus giving a less steep slope.

Now consider a more complex example:

Since we cannot find the probability for a specific value, we will consider the

probabilities for two intervals, from x 1 to x 1+ ∆ and from x 2 to x 2+ ∆ . Notice that we are

taking intervals of the same length, just starting at different points, and that ∆ is a

very small value.

p1=P [ x 1< X ≤ x 1 +∆ ]=F X ( x1 + ∆ )−F X ( x 1 )

p2=P [ x 2< X ≤ x 2 +∆ ]=F X ( x 2+ ∆ ) −F X ( x 2 )

Even though we are taking intervals of the same length, the probability values we

find for those intervals are greatly different. This is because the density of

probabilities in the latter interval is higher. Thus, the probability of an interval is

proportional to the probability density in that interval. This proportionality becomes

an equality when we multiply the probability by the length of the interval.

Another indication of the density is the steepness of the CDF curve. A flat area

indicates a lower density, while a steep area indicates a higher density and vice

versa.

The slope of the CDF curve is of course given by its first derivative, the rate of

change. For a specific value of x ,

p=P [ x < X ≤ x + ∆ ]
¿ F X ( x+ ∆ )−F X ( x )

F X ( x +∆ )−F X ( x )
¿ ⋅∆
∆

CDF values simply indicate a probability. Unlike PMFs, they indicate the probability

over a range, but the value is still a probability. Thus, the first term in the equation

above can be thought of as the probability in an interval divided by the length of the

interval, the average probability density. From this, we can go back to finding the

total probability by multiplying it with the length of the interval.

If we use very small values of ∆ , we could even go as far as to say the probability at

the single point x is approximately equal to the probability density for the interval

from x to x +∆ . Of course, this approximation will be extremely inaccurate if we use

F x ( x +∆ )−F x ( x )
large values of ∆ . If ∆ → 0, lim represents the probability density near
∆→0 ∆
x . This probability density near x is denoted by f X ( x ). Thus,

F x ( x +∆ )−F x ( x ) d
f X ( x )=lim = F X ( x)
∆→0 ∆ dx

This is the PDF of X . The PDF is the probability density per unit length for a random

variable X near the point x .

Always remember that the PDF is not a probability, but rather a probability density. By

the axioms of probability, the probability at a point must be between 0 and 1.

However, the probability density at a point can be more than 1, if there is a high

probability within a small interval.

PDF Curves

The PDF curve understandably depends on the CDF values. For the case where

probabilities are uniformly distributed over an interval, the PDF curve will be flat.

Of course, the area of the PDF curve for the interval will represent the probability in
1
that interval. As such, the total area will be 1. This means the height of the line is
b−a
.

For non-uniform distributions, the PDF curve will be curved.

Thus, P [ a< X ≤ b ] is given by the area of the curve, which, in this case, can be
b

calculated by ∫ f X ( x ) dx. If a=−∞ , this just gives us F X ( b ) .

a
 {
0 x <0

{
1
x 0< x <4
We previously saw an example where F X ( x )= 0< x < 4 . Here, f X ( x )= 4 .
4
0 otherwise
1 x≥4
x
1 x
Inversely, ∫ dx= .
−∞ 4 4

+∞

Another thing to note is that ∫ f X ( x ) dx=1.

−∞

Of course, the PDF formulas are not always simple, and since integration and

differentiation have become involved, we will soon start facing problems like having

to integrate by parts.
Properties of PDFs

+∞
 ∫ f X ( x ) dx=1
−∞

x
 ∫ f X ( x ) dx=P [ X ≤ x ] =F X ( x )
−∞

b
 P [ a< x ≤ b ] =∫ f X ( x ) dx
a

 f X ( x )≥ 0

Properties of CDFs

x
 F X ( x )=P [ X ≤ x ] =∫ f X ( x ) dx
−∞

 Continuous function, and thus differentiable

 F X (−∞ )=0
 F X ( +∞ )=1
 P [ a< x ≤ b ] =F X ( b )−F X ( a )

Example

{
−x
2
f X ( x )= cx e x≥0
0 otherwise

We need to find

i) Find the value of c ,

ii) Derive F X ( x ),
 iii) Retrieve the PDF from the CDF, although it is already given,

iv) Find P [ a< x ≤ 5 ] using the PDF and the CDF separately.

This problem involves integration by parts ( ∫ u ⋅dv =u ⋅v− ∫ v ⋅du ).

i)

+∞

∫ f X ( x ) dx=1
−∞

+∞ −x

∫ cx e 2
dx=1
0

[cx (−2 ) e ]
−x +∞ +∞ −x
2
0 −c ∫ 1 ⋅ (−2 ) e 2
dx=1
0

[ ]
−x + ∞
2
0+2 c (−2 ) e 0 =1

4 c=1 c
1
c=
4

ii)

x
F X ( x )=∫ f X ( x ) dx
−∞

x −x
1
¿∫ xe 2
dx
0 4
 [ ]
−x x −x
1 1
¿ x (−2 ) e
4
2
−
4
∫ 1 ⋅ (−2 ) e 2
dx
0

[ ]
−x −x x
1 1
¿− x e 2 + (−2 ) e 2
0
2 2
−x −x
1
¿− x e 2 −e 2 +1
2
−x −x
1
¿ 1−e 2 − x e 2
2

iii)

d
f X ( x )= F (x )
dx X

( )
−x −x
d 1
¿ 1−e 2 − xe 2
dx 2

[ ]
−x −x −x
1 1 1
¿ 0+ e 2 − e 2 − x e 2
2 2 2

−x
1 2
¿ xe
4

iv)

Using CDF,

P [ 2< X ≤5 ] =F X ( 5 )−F X ( 2 )

( )( )
−5 −5 −2 −2
1 1
¿ 1−e 2 − ⋅5 ⋅e 2
− 1−e 2 − ⋅2 ⋅e 2
2 2
 −5
−1 7 2
¿2e − e
2
Using PDF,

5
P [ 2< X ≤5 ] =∫ f X ( x ) dx
2

5 −x
1
¿∫ xe 2
dx
2 4

[ ]
−x 5 5 −x
1 1
¿ x (−2 ) e 2
− ∫ 1⋅ (−2 ) e 2
dx
4 2 4 2

[ ]
−5 −x 5
5
¿− e 2 + e−1− e 2
2
2

−5
7
¿ 2 e−1− e 2
2
Expectations and Variances

For discrete random variables, we defined the expectation as

E [ X ]= ∑ x ⋅ P X ( x )
x∈ SX

For continuous random variables, it is defined as

+∞
E [ X ] = ∫ x ⋅ f X ( x ) dx
−∞

For discrete random variables, we defined variance as

Var [ X ] = ∑ ( x −E [ x ])2 ⋅ P X ( x )
( x ∈S X )

For continuous random variables, it is defined as

+∞
Var [ X ] = ∫ ( x−E [ x ]) ⋅ f X ( x )
2

−∞

Note that the limits for integration are replaced by the given limits for individual

problems. ∞ has been used for show here.

Well-Known Continuous Random Variables

Uniform Continuous Random Variables

We have already seen a uniform random variable for a discrete set of values. In the

continuous version of this, the values are given by a set [ a , b ]. This set defines the

possible values of X .

X =uniform continuous random variable∈the interval [ a , b ]

X uniform ( a , b )

Note that the brackets do not matter for continuous random variable and can be

used interchangeably.

Unlike discrete random variables, we cannot simply define the uniform continuous

random variable as one that has an equal probability for every possible value. This is

because all values for continuous random variables always have the value 0 , as we

have previously discussed. As such, there are two possible definitions:

i) X is a uniform continuous random variable if the probability density is

constant

ii) X is a uniform continuous random variable if it is equally likely that a point

from any interval within the specified range will be picked. This definition is

a little more confusing and is less preferred.

By this definition, the PDF can be defined as:

f X ( x )= {c0 a< x ≤ b
otherwise

We know,

+∞

∫ f X ( x ) dx=1
−∞

∫ c dx=1
a

[ cx ] ba=1

c ( b−a )=1

1
c=
b−a

Thus, the PDF is given by

{
1
a< x ≤b
f X ( x )= b−a
0 otherwise
Using this, we can calculate the CDF as

x
1
F X ( x )=∫ dx
a b−a

[ ]
x
x
¿
b−a a

x−a
¿
b−a

{
0 x <a
x−a
F X ( x )= a< x ≤ b
b−a
1 x≥b

For uniform random variables, the expectation is given by

 a+ b
E [ X ]=
2

and the variance is given by

( b−a )2
Var [ X ] =
12
Example

Say a student requires between 22 to 30 minutes to travel from home to school. The

probability density between this interval is constant. Say the class starts at 8 :00 AM

and the student starts from home at 7 :35 AM . What is the probability that they reach

in time?

Let X be a uniform continuous random variable, where X =amount of delay . The best

possible case is that 22 minutes are needed, which puts the arrival time at 7 :57 AM ,

meaning X =−3. The worst possible case is 30 minutes are needed, which puts the

arrival time at 8 :05 AM , meaning X =5. Thus, X is uniformly distributed between

[ −3 , 5 ].

{
1
−3 ≤ x ≤ 5
f X ( x )= 8
0 otherwise

0
1
P [−3≤ X ≤ 0 ] =∫ dx
−3 8

[]
0
x
¿
8 −3

3
¿
8

An easier way we could have calculated this probability is if we defined the ‘length’ of

valid values as A , from 7 :57 AM to 8 :00 AM , and the ‘length’ of the total range as B,

from 7 :57 AM to 8 :05 AM . Thus,

 A 3
P [3 ≤ X ≤ 5 ]= =
B 8

Exponential Random Variable

Exponential random variables are related to Poisson distributions. Consider a

scenario where events are occurring at a particular Poisson rate, λ . Say n events

occur in a time period T .

Let X 1 =time of first event , X 2 =timebetween second event ∧first event and so on. This can

be generalized to

X i =time between ( i−1 )−th∧i−th event where i≥ 2

Since the events could occur at absolutely any moment, all the X we just defined are

continuous random variables. For exponential distributions, we want to find the

distribution, i.e. the CDF and PDF, of X 1.

We know that F X ( x )=P [ X 1 ≤ x ]. Instead of finding this, we will instead find

F X ( x )=1−P [ X 1 > x ] =1−F X ( x ), the complimentary CDF. The complimentary CDF tells
C

us the probability that the first event occurs after the time x . This is the same as the

probability that there is no event between time ( 0 , x ).

P [ 0 arrivals∈time ( 0 , x ) ] counts the number of arrivals in an interval. We know that this

is defined by the Poisson distribution. Thus

|
− λx
C e ( λx )n
F ( x )=
X
n! n=0

−λ x
e ( λx )0
¿
0!

− λx
¿e

From this,

F X ( x )= { 0
1−e − λx
x <0
x≥0

We can also find the PDF from this.

d
f X ( x )= F (x )
dx X

− λx
¿ λe
 {
− λx
f X ( x )= λ e x≥0
0 otherwise

The expected value of an exponential random variable is given by

1
E [ X ]=
λ

and the variance is given by

1
Var [ X ] = 2
λ

The expected value actually tells us something very obvious in this case. If there are
1
5 events per hour, λ=5. Thus, the expected interval time will be hours, or 12
5
minutes, meaning on average, there will be an arrival every 12 minutes. Note that this

is on average. In reality, the arrivals are random.

Exponential random variables can be related to geometric random variables, since in

geometric random variables we were interested in the number of Bernoulli

experiments before the first success and here we are interested in the time before

the first event.

Example

1
On an average, there is an earthquake every 3 months. Thus, =3 . We need to find
λ
the probability that the next earthquake will occur after 3 months, but before 7

months.

Say X =time ¿ next earthquake , meaning X exponential ( λ ). Whenever we are given

problems related to some rate, if we have to work with the rate of events occurring

then we want a Poisson random variable, and if we have to work with the time interval

between events, then we want an exponential random variable.

P [ 3 < X ≤7 ] =F X ( 7 )−F X ( 3 )

−7 −3
¿ 1−e 3 −1+ e 3

¿ 0.27
Gamma Random Variables

Gamma distributions are a generalization of the exponential distribution, and as

such, are also related to Poisson distributions.

Say X =time ¿ get k−th arrival= X 1+ X 2 +…+ X k.

In this case, the PDF is given by

{
λ ( e ( λx ) )
−λx k−1

f X ( x )= x≥0
( k −1 ) !
0 otherwise

Clearly, this becomes the same as the exponential distribution for k =1.

Also, since we have a factorial in the denominator, k has to be an integer.

There are situations where k is a real number, and in those cases, we need to

convert this into a gamma function.

∞
Γ ( α )=∫ x
α −1 − x
e dx
0

Γ ( 1 )=1

Γ ( 2 )= (2−1 ) ⋅ Γ (1 )=1⋅ 1=1!

Γ ( 3 ) =( 3−1 ) ⋅ Γ ( 2 )=2 ⋅1=2 !

Γ ( 4 )= ( 4−1 ) ⋅ Γ (3 )=3 ⋅2 ⋅1=3 !

Thus, in general we can say

Γ ( r +1 ) =r !

and

{
λ(e ( λx )r−1 )
−λx

f X ( x )= x≥0
Γ (r )
0 otherwise

The gamma distribution curves are different from other distributions, due to the

existence of two parameters, X gamma ( r , λ ). Thus, the graph depends on which

parameter we keep fixed.

We do not need to remember these graphs.

The expected value of a gamma distribution is given by

r
E [ X ]=
λ

and the variance is given by

r
Var [ X ] = 2
λ

Gamma distributions are related to negative binomial distributions, since in negative

binomial distributions, we are interested in the number of attempts required for k

successes, and here we are interested in the time until the k -th event.

Example

Say customers are arriving at a restaurant at a rate of 12 customers per hour and

that the restaurant will start having profits after 30 customers have arrived. We want

to find the expected time until this happens.

Let X =time ¿ make profits , where X gamma ( 31 ,12 ) .

31
E [ X ]= =2 hours 35 minutes
12
Gaussian Random Variables

Also called the normal random variable, the gaussian random variable is possibly the

most important random variable due its wide usage in statistical inferences.

Whenever we are unsure about what the random variable should be in a situation, it

is best to use the gaussian random variable.

The PDF is given by

2
− ( x−μ )
1
, where −∞ < x <+∞ .
2
2σ
f X ( x )= e
√2 π σ 2

The two parameters are σ 2, which is the variance of the random variable and μ, which

is the expectation of the random variable.

Notice that the curve is symmetric, and if we change the value of μ, the only thing

that happens is that the curve shifts. The shape of the curve is unaffected. However,

increasing the value of σ 2 will flatten the curve.

Note that the width of the curve has to increase if the height decreases, so as to

maintain the area under the curve at 1.

The problem with gaussian random variables is that, if we integrate the PDF, we do

not get a closed-form solution. As such, we cannot mathematically define a value for

the CDF. This means we cannot find the probability associated with a particular

interval using gaussian random variables.

The probabilities in these scenarios are calculated using numerical techniques that

we will not cover in this course. Another option is to use tables and the PDF curve.

For any given curve,

P [ μ−σ < X < μ+σ ] =0.68

P¿

P [ μ−3 σ < X < μ+3 σ ]=0.997

We can use tables or computer applications or some specific calculators to calculate

the CDF values for Gaussian random variables. Here, we shall be looking into how to

use those tables.

X N ( μ , σ 2 ) where μ is the mean and −∞ < μ<+ ∞ and σ 2 is the variance and σ 2> 0

The table should contain all the CDF values for each pair of μ and σ 2. There could be

millions of values like this, and it would be unrealistic to expect a table to include all of

the values.

Because of this limitation, we will be considering a special gaussian random variable,

Z ( 0 , 1 ). Such a gaussian random variable, which has a 0 mean and a 1 variance, is

called a standard normal random variable.

For Z ,

2
−x
1
f X ( x )= e 2

√2 π

Normally, we would denote the CDF of a random variable as F X ( x ), but the CDF of Z

uses a special character, Φ.

Φ ( z )=P [ Z ≤ z ]
In most cases, the value of z in tables is between 0 to 4 . The table is setup in a weird

way to save space. Each for indicates the values for some values of z and the column
1
specifies the digit in the ths place of z . Thus, to find the CDF for z=1.26 , we need
100
to go to the row marked 1.2 and the column marked 0.06 , where we find the value
0.896 .

However. there are no probability values available for negative values of z in the

table. For negative values, we use the symmetric property of the curve.

Each value from the table tells us the area under the curve from the leftmost point of

the curve to the specified value of z . Say we are asked to find P [ Z >1.26 ]. This is just
1−0.896=0.104 . We took the area on the left of z=1.26 and subtracted it from 1 to

get the area on the right of z=1.26 .

Now, using the symmetric property of the graph, we can tell that the area on the

right of z=1.26 is the same as the area on the left of z=−1.26 .

Thus, P [ Z ←1.26 ] =P [ Z >1.26 ] =1−P [ Z <1.26 ]=1−0.896=0.104 .

Following a similar pattern of thinking, P [ Z >−1.26 ] =P [ Z<1.26 ] =0.896 .

A few more cases are:

 P [− z< Z < z ]
¿ P [ Z < z ] −P [ Z ← z ]

¿ Φ ( z )−( 1−Φ ( z ) )

¿ 2 Φ ( z )−1

 P [ Z > z ] =0.05
1−P [ Z ≤ z ]=0.05

P [ Z ≤ z ] =1−0.05

P [ Z ≤ z ] =0.95

z=1.65

We are approximating from the table since the exact value is not available in

given table. There are two possible values (1.64 and 1.65) and one has been

chosen that felt more accurate. If a more accurate table were available that

had more accurate values of z , the exact value could be found.

 P [− z< Z < z ] =0.90

P [ Z < z ] −P [ Z ← z ] =0.90

P [ Z < z ] −( 1−P [ Z < z ] )=0.90

2 × P [ Z < z ] =1.90

P [ Z < z ] =0.95

z=1.65
Now to get back to non-standard normal random variables, i.e. X N ( μ , σ 2 ) where μ ≠ 0

and σ 2 ≠ 1. For these cases, we can do two things. We could either have this huge

table for all possible values of μ and σ 2, or we could have some formula that allows us

to convert a normal random variable into a standard normal random variable. This

formula is

X−μ
Z=
σ

Say Y is a random variable that is derived from X . Any random variable that is derived

from a gaussian random variable is also a gaussian random variable. Thus,

Y Gaussian ( μ1 , σ 21 )

For a derived random variable where Y =aX +b , two formulae we need to remember

are:

E [ Y ] =a ⋅ E [ X ] +b

Var [ Y ] =a Var [ X ]
2

Using these formulae,

μ1=E [ Y ] =E
[ σ ]
X −μ E [ X ] −μ μ−μ
=
σ
=
σ
=0

1 1 2
Var [ X ] = 2 ⋅ σ =1
2
σ 1= 2
σ σ

X −μ
Thus, we have standardized Y by using the formula Y = .
σ
Example

Say some data packets we are sending have an average delay that is a normal

distribution with μ=10 and σ 2=4 . We want to find the probability that the delay for a

specific packet is greater than 13.

Let X be a random variable such that X =average delay of a packet , i.e. X N ( 10 , 4 ).

∴ P[ X >13]

¿P
[ X−μ 13−μ
σ
>
σ ]
[ ]
¿ P Z>
3
2

¿ 1−P [ Z ≤1.5 ]

¿ 1−0.933

¿ 0.067
Beta Random Variables

For beta random variables, the PDF is defined as

1 α−1 β −1
f X ( x )= x ( 1−x ) 0< x <1
B (α , β )

Here,

1
B ( α , β )=∫ x
α −1 β−1
( 1−x ) dx
0

Γ (α) Γ ( β )
¿
Γ ( α+β)

Beta distributions are used for calculating probabilities that are fractions.

For example, when send data packets, what is the portion of packets that were

successfully sent? (I have absolutely no clue what this means.)

The graphs of beta distributions are complex and change depending on the values of
β and α .

Beta distributions will be covered in far more details in later chapters.