Analysis of Production Systems (4DC10)

Week 1 (Probability models)

Michel Reniers

Department of Mechanical Engineering

Probability models

Relevant in many areas, such as:

• Manufacturing
• Computer networks (Internet)
• Supply chains
• Statistical physics
• Quantum physics
• Random heterogeneous materials
• Biology
• Computer simulation

Learning goal: Be able to formulate and analyze basic probability models

2
Magician

A magician takes a random card from a deck of 52 cards. You have to guess which card
he is holding.

Before guessing you may ask one of the following two questions:

1. Is the card a black one (spades or clover)?

2. Is the card jack of hearts?

Question: Which question will maximize the probability of guessing the right card?

A Question 1
B Question 2
C Does not matter
3
Magician (Answer)
• “Is the card a black one?”

With probability 21 the answer is “yes”. Then 26 cards remain to choose from, so
1
your probability of guessing the right one is 26 . A similar reasoning applies for
answer “no”.
1 1 1 1 1
The probability of guessing the right one is 2 · 26 + 2 · 26 = 26

4
Magician (Answer)

• ‘Is the card jack of hearts?”.

1
With probability 52 the answer is yes and you are done.

51
With probability 52 the answer is no, and then the probability of guessing the right
1
one is 51 .

1 51 1 1
So the probability of guessing the right one is 52 ·1+ 52 · 51 = 26
4
Magician (Answer)
• “Is the card a black one?”

With probability 21 the answer is “yes”. Then 26 cards remain to choose from, so
1
your probability of guessing the right one is 26 . A similar reasoning applies for
answer “no”.
1 1 1 1 1
The probability of guessing the right one is 2 · 26 + 2 · 26 = 26
• ‘Is the card jack of hearts?”.
1
With probability 52 the answer is yes and you are done.

51
With probability 52 the answer is no, and then the probability of guessing the right
1
one is 51 .

1 51 1 1
So the probability of guessing the right one is 52 ·1+ 52 · 51 = 26
4
Monty Hall dilemma

• It is the climax of a game-show.

• You have to choose a door out of three, behind one of them is the car of your
dreams and behind the others a goat.
• You choose a door without opening it. The host (knowing what is behind the
doors) then opens one of the remaining doors, showing a goat.

Question: Now you are given the opportunity to switch doors: Are you going to switch?

A Yes
B No
C Why bother, there is no difference

5
Monty Hall dilemma (Answer)

6
Our intuition on probability: Empirical law of large numbers
• Throwing a fair coin: Fraction of Heads should be 1
2 in the long run

• Relative frequency of event H (Head) in n repetitions of throwing a coin is

n(H)
fn (H) =
n
where n(H) is number of times Head occurred in the n repetitions

• Then
1
fn (H) → as n → ∞.
2

• In general: The relative frequency of event E approaches a limiting value as the

number of repetitions tends to infinity.

• Intuitively we would define the probability of event E as this limiting value.

7
Ingredients of a probability model

Sample space S can be discrete or continuous: Events are subsets of S:

• Flipping a coin, S = {H, T} E = {H}
• Rolling a die, S = {1, 2, ... , 6} E = {1, 2}
• Rolling a die twice, S = {(i, j) | i, j = 1, 2, ... , 6} E = {(1, 2), (3, 4), (5, 6)}
• Process times, S = [0, ∞) E = (0, 1)
• Number of machine failures, S = {0, 1, 2, ...} E = {3, 4, ...}
• Throwing darts, S = {(x, y) | x2 + y 2 ≤ 1} E = {(x, y), 0 ≤ x ≤ 1
4, 0 ≤ y ≤ 41 }
p

Probability measure P : 2S → [0, 1] that assigns a probability to events

8
Probabilities on Events

For each event E there is a number P(E) (the probability of E) such that:
• 0 ≤ P(E) ≤ 1
• P(S) = 1
• E1 , E2 , ... mutually disjoint (have nothing in common), then
P(E1 ∪ E2 ∪ · · · ) = P(E1 ) + P(E2 ) + · · ·

E1 E2 E3 E4

9
Probabilities on Events

For each event E there is a number P(E) (the probability of E) such that:
• 0 ≤ P(E) ≤ 1
• P(S) = 1
• E1 , E2 , ... mutually disjoint (have nothing in common), then
P(E1 ∪ E2 ∪ · · · ) = P(E1 ) + P(E2 ) + · · ·

Examples:
• Flipping a coin, P({H}) = P({T}) = 1
2
• Rolling a die, P({1}) = 61 , P({1, 2}) = P({1}) + P({2}) = 1
3
• Rolling a die twice, P({(i, j)}) = 1
36
• Throwing darts, P(E) = area of E, divided by area of unit disk

10
Probabilities on Events

For each event E there is a number P(E) (the probability of E) such that:
• 0 ≤ P(E) ≤ 1
• P(S) = 1
• E1 , E2 , ... mutually disjoint (have nothing in common), then
P(E1 ∪ E2 ∪ · · · ) = P(E1 ) + P(E2 ) + · · ·

Examples:
• Discrete
PS, assign p(s) ≥ 0 for each s ∈ S such that p(s) = 1. Then
P
s∈S
P(E) = s∈E p(s)
• Equally likely outcomes s1 , ... , sN , so p(si ) = 1/N and
|E|
P(E) =
N
11
Some basic rules that follow from the ingredients

• If E ⊂ F, then P(E) ≤ P(F)

• If finitely many E1 , E2 , ... , En are mutually disjoint, then
P(E1 ∪ E2 ∪ · · · ∪ En ) = P(E1 ) + P(E2 ) + · · · + P(En )

• P(E c ) = 1 − P(E) where E c is the complement of E, so E c = S \ E

• P(E ∪ F) = P(E) + P(F) − P(EF) where EF = E ∩ F

E EF F

12
Theoretical law of large numbers
• Flipping a coin an unlimited number of times, then an outcome is an infinite
sequence s of Heads and Tails,
s = (H, T, T, H, H, H, T, ...).

• Let Kn (s) denote the number of Heads in the first n flips of outcome s.
• Then according the theoretical (strong) law of large numbers,
Kn (s) 1
lim =
n→∞ n 2
with probability 1
• In general: If an experiment is repeated an unlimited number of times, and if the
experiments are independent of each other, then the fraction of times event E
occurs converges with probability 1 to P(E).
• The method of computer simulation is based on this law!

13