Continuous distributions

Normal distribution

X ⇠ N(µ, 2 ), with expectation µ and variance 2 . The probability

density function is
✓ 2
◆
1 (x µ)
pX (x) = p exp
2⇡ 2 2 2

X µ a µ
To calculate P(X < a): standardize it Z = , and find P Z <
from the standard normal table.
4
The standard normal

• In general, it is hard to calculate P(Z < a) for any a even if Z is a

standard normal distribution.
• People make a standard normal table to have quick reference

• It gives you value about P(Z < a), for example

P(Z < 0.21) = 0.5832, P(Z < 0.33) = 0.6293

Each row indicates the first digit, columns for the second digit.

5
Another way to view normal

You can do the opposite of standarization, to create a general normal

distribution from the standard normal.
Suppose Z ⇠ N(0, 1), let

X =µ+ ⇥Z

Then
2
X ⇠ N(µ, )
2 2
Why? Remeber E [X ] = µ + ⇥ E [Z ] = µ and Var[X ] = Var[Z ] = .
Linear transformation keeps normal distribution.

6
More calculations of normal distribution

Suppose the time of people using personal computers for entertainment

per day is normal distributed with mean 2 hours and standard deviation
0.5 hours. X ⇠ N(2, 0.52 )
• Find prob the entertainment hours is between 1.8 and 2.75 hours per
day.

Note that P(1.8  X  2.75) = P(X  2.75) P(X  1.8)! It’s

the shaded area under the curve. 7
More calculations of normal distribution

P(1.8  X  2.75) = P(X  2.75) P(X  1.8)

Read from the standard normal table directly, we have
P(X  2.75) = P(Z  2.75 2
0.5 ) = P(Z  1.5) = 0.9332
1.8 2
Next P(X  1.8) = P(Z  0.5 ) = P(Z  0.4)
However, standard normal table only shows probability of positive values.
Again, use a little bit geometry
P(Z  0.4) = P(Z 0.4) = 1 P(Z  0.4) = 1 0.6554 = 0.3446
Therefore P(X  2.75) = 0.9332 0.3446 = 0.5886

8
Standard normal table

x 0.00 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09

0.0 0.5000 0.5040 0.5080 0.5120 0.5160 0.5199 0.5239 0.5279 0.5319 0.5359
..........
0.3 0.6179 0.6217 0.6255 0.6293 0.6331 0.6368 0.6406 0.6443 0.6480 0.6517
0.4 0.6554 0.6591 0.6628 0.6664 0.6700 0.6736 0.6772 0.6808 0.6844 0.6879
..........
1.4 0.9192 0.9207 0.9222 0.9236 0.9251 0.9265 0.9279 0.9292 0.9306 0.9319
1.5 0.9332 0.9345 0.9357 0.9370 0.9382 0.9394 0.9406 0.9418 0.9429 0.9441
1.6 0.9452 0.9463 0.9474 0.9484 0.9495 0.9505 0.9515 0.9525 0.9535 0.9545
1.7 0.9554 0.9564 0.9573 0.9582 0.9591 0.9599 0.9608 0.9616 0.9625 0.9633
1.8 0.9641 0.9649 0.9656 0.9664 0.9671 0.9678 0.9686 0.9693 0.9699 0.9706
1.9 0.9713 0.9719 0.9726 0.9732 0.9738 0.9744 0.9750 0.9756 0.9761 0.9767
2.0 0.9772 0.9778 0.9783 0.9788 0.9793 0.9798 0.9803 0.9808 0.9812 0.9817
..........
9
Approximate binomial by normal
1
DeMoivre first proved the approximation with p = 2 in 1733, then
Laplace extended to general p in 1812.

10
Approximate binomial by normal

The idea size of a first-year class at a college is 150 students. The college
knows that on average 30% students accept the o↵er. The college makes
450 o↵ers this year. Calculate the probability of more than 150 students
accepting the o↵er.
Let X denote number of students accept the o↵er, it follows the binomial
distribution Bin(450, 0.3), we want to calculate P(X > 150). We
approximate it by normal N(np, np(1 p))

✓ ◆
X 450 ⇥ 0.3 150 450 ⇥ 0.3
P(X > 150) = P p >p = P(Z > 1.59)
450 ⇥ 0.3 ⇥ 0.7 450 ⇥ 0.3 ⇥ 0.7

Search the standard normal table, we know P(Z < 1.59) = 0.9441. So
the result is 0.0559.

11
Exponential distribution

How to model the amount of time until something happen

• the next email arrival

• an accident happens
• a light bulb burns out
• Notation: X ⇠ Exp( )

Similar to geometric distribution (discrete!)

The exponential distribution is used to model waiting time. We will see
application of it in inventory management later.

12
Exponential distribution

An exponential random variable has probability density function

(
e x x 0
fX (x) =
0 otherwise

Cumulative distribution function

(
x
1 e x 0
FX (x) =
0 otherwise

E [X ] = 1 , Var[X ] = 1
2 , E [X n ] = n E [X n 1
]

13
Exponential distribution

Left: PDF, Right: CDF of exponential distribution Exp(1).

14
Example

Suppose that the length of a phone call in minutes follows exponential

1
distribution with = 10 . If you see someone in the public telephone
booth, find the probability you have to wait

• more than 10 minutes.

1
P(X > 10) = 1 P(X  10) = 1 F (10) = e = 0.368

• between 10 and 20 minutes.

P(10  X  20) = P(X  20) P(X  10)

2 1
= (1 e ) (1 e )
= 0.233

15
Memoryless of exponential

• X ⇠ Exp( )
• P(X > s + t | X > s) =?
• Remeber that the CDF is FX (x) = 1 e x

•
P(X > s + t and X > s)
P(X > s + t | X > s) =
P(X > s)
P(X > s + t)
=
P(X > s)
(s+t)
e
= s
e
t
=e = P(X > t)

16
Example of the memoryless property

You enter a post office, there are two clerks serving two customers.
Suppose a clerk serving time for any customer is exponential distribution
. You will be served once any of the two customers finish. What’s the
probability that you are the last customer to leave the post office?

• When you start being served, one customer (say customer A) has
already left the post office.
• Because of the memoryless property, the serving time for you and
customer B are the same exponential distribution with .
• By symmetry, you and customer B have the same distribution of
serving time, so you have equal probability of finishing first. 12 .

17
Memoryless of exponential

What are the implications of the memoryless property?

• Exponential is the unique continuous distribution with memoryless

property.
• If you are waiting for a bus, assume your waiting time follows exp.
conditional on you have been waiting for 30 minutes, the distribution
simply forgets everything in the past. Your remaining waiting time
is the same as just coming to the bus stop.
• If the lifetime of a machine follows exp distribution, no matter how
long this machine has been functional, conditional on living very
long, the machine is as good as new. No wear-and-tear e↵ect.
• If human life follows exp distribution, a 80 years old man’s remaining
lifetime is the same as a newborn baby!

Clearly, it is not appropriate for human or machine life.

18
Memoryless of exponential

Why / when do we use it?

• Some physical phenomena such as radioactive decay truly has

memoryless property. So exponential distribution is an important
model in its own right.
• It is a building block of many useful stochastic process for inventory
management. If we assume each customer acts independently. You
may expect the arrival of next customer’s order follows exponential
distribution.
• We will revisit exponential distributions later in queue theory.

19