Probability

 Where do probabilities come from?

 Conditional probability: the art of
asking the right question
 Calculating probabilities from tables
 Introduction to R

STA 309 – Business Statistics

September 1, 2020
Dr. Shannon Provost
Department of Information, Risk, & Operations Management
 Probability is a science
of uncertainty.

https://www.visualcapitalist.com/measuring-perceptions-of-uncertainty/
Probability notation

 The probability of some event is a number between zero and one.

0 ≤ 𝑃(𝑠𝑜𝑚𝑒 𝑒𝑣𝑒𝑛𝑡) ≤ 1
 An event with probability 1 is certain to occur; an event with
probability 0 is certain not to occur.
 For example:
 𝑃 𝑐𝑜𝑖𝑛 𝑙𝑎𝑛𝑑𝑠 ℎ𝑒𝑎𝑑𝑠 = 0.5
 𝑃 𝑓𝑙𝑖𝑔ℎ𝑡 𝑑𝑒𝑝𝑎𝑟𝑡𝑠 𝑜𝑛 𝑡𝑖𝑚𝑒 = 0.79
 𝑃 𝑟𝑎𝑖𝑛 𝑖𝑛 𝐴𝑢𝑠𝑡𝑖𝑛 = 0.2
 𝑃 𝑟𝑎𝑖𝑛 𝑖𝑛 𝐷𝑢𝑏𝑙𝑖𝑛 = 0.4
 𝑃 𝑐𝑜𝑙𝑑 𝑑𝑎𝑦 𝑖𝑛 𝐻𝑒𝑙𝑙 = 0.000000000000001
Where do probabilities come from?

 Careful counting
 Data
 Subjective judgment
 Other probabilities
Where do probabilities come from?

 Careful counting
 Data
 Subjective judgment
 Other probabilities
Example: draw a card at random
from a standard 52-card deck

What is 𝑃 𝑑𝑖𝑎𝑚𝑜𝑛𝑑 ?

What is 𝑃 𝑘𝑖𝑛𝑔 𝑜𝑟 𝑞𝑢𝑒𝑒𝑛 ?

 13
𝑃 𝑑𝑖𝑎𝑚𝑜𝑛𝑑 =
52
 13
𝑃 𝑑𝑖𝑎𝑚𝑜𝑛𝑑 =
52
 8
𝑃 𝑘𝑖𝑛𝑔 𝑜𝑟 𝑞𝑢𝑒𝑒𝑛 =
52
 8
𝑃 𝑘𝑖𝑛𝑔 𝑜𝑟 𝑞𝑢𝑒𝑒𝑛 =
52
The counting principle:

# 𝑜𝑢𝑡𝑐𝑜𝑚𝑒𝑠 𝑜𝑓 𝑖𝑛𝑡𝑒𝑟𝑒𝑠𝑡 𝑌
𝑃= =
𝑡𝑜𝑡𝑎𝑙 # 𝑜𝑓 𝑜𝑢𝑡𝑐𝑜𝑚𝑒𝑠 𝑁

Key assumption: All outcomes equally likely

Example: rolling two fair dice

Let 𝑆 be the sum of the two numbers.

Clearly 𝑆 is random: it varies

from one roll to the next.

What is 𝑃 𝑆 ≥ 9 ?
 10
𝑃 𝑆≥9 = ≈ 0.28
36
 10
𝑃 𝑆≥9 = ≈ 0.28
36
Where do probabilities come from?

 Careful counting
 Data
 Subjective judgment
 Other probabilities
𝑃 𝑛𝑒𝑤𝑏𝑜𝑟𝑛 𝑏𝑎𝑏𝑦 𝑖𝑠 𝑎 𝑔𝑖𝑟𝑙
100
=
206
≈ 0.485
 Joe will be in a
P ≈ 0.009
car accident this year
 Emma Pierson & Stanford colleagues analyzed racial bias across
100 million traffic stops
 Data set incorporates 21 state patrol agencies and 35 municipal police
departments over 10 years

 Goal: test the Grogger & Ridgeway “veil of darkness” hypothesis

 Prediction: the probability that a stopped driver is Black will be
lower at night, when it’s harder to see the race of the driver.
 An illustration of the veil-of-darkness test for stops
occurring in three short time windows in Texas.

Each panel shows a different time for the onset of dusk (end of civil twilight).
Where do probabilities come from?

 Careful counting
 Data
 Subjective judgment
 Other probabilities
 𝑊𝑖𝑙𝑙𝑖𝑎𝑚𝑠 𝑤𝑖𝑛𝑠
P 𝑎𝑛𝑜𝑡ℎ𝑒𝑟
𝑊𝑖𝑚𝑏𝑙𝑒𝑑𝑜𝑛

= 0.3 ?
= 0.2 ?
𝑃 = 0.2 𝑃 = 0.3
 Give me $10.
If Serena wins another Wimbledon,
I’ll repay your $10…and I’ll pay you
another $30.
Expected payoff:

𝐸 = 10 ∗ 0.8 − 30 ∗ 0.2
=8−6
=2
 Expected payoff:

𝐸 = 30 ∗ 0.3 − 10 ∗ 0.7
=9−7
=2
OK!
𝑃 𝐴𝑝𝑝𝑙𝑒 𝑠𝑡𝑜𝑐𝑘 𝑢𝑝 𝑛𝑒𝑥𝑡 𝑦𝑒𝑎𝑟 = 0.7 ? ≈(Buy!)
0.28

𝑃 𝐴𝑝𝑝𝑙𝑒 𝑠𝑡𝑜𝑐𝑘 𝑢𝑝 𝑛𝑒𝑥𝑡 𝑦𝑒𝑎𝑟 = 0.4 ? ≈(Sell!)

0.28
Where do probabilities come from?

 Careful counting
 Data
 Subjective judgment
 Other probabilities
Where do probabilities come from?

 Careful counting
 Data
 Subjective judgment
 Other probabilities
Negation Rule

 In general, for any event 𝐴, 𝑃 𝑛𝑜𝑡 𝐴 = 1 − 𝑃(𝐴)

 Also known as the “complement rule”
 Let’s pretend that the probability of rain tomorrow is 0.3
 Then clearly the probability that it will NOT rain tomorrow is 0.7

 Say the probability that Tiger Woods wins the Masters

tournament in November is 0.13.
 Then the probability that Tiger Woods does not win the Masters is 0.87.
Addition Rule

 Draw a card at random. What is 𝑃 𝑐𝑎𝑟𝑑 𝑖𝑠 𝐴𝑐𝑒 𝑜𝑟 𝑆𝑝𝑎𝑑𝑒 ?

 Key insight: “Ace” and “Spade” are not mutually exclusive outcomes.
 Therefore we cannot simply add 𝑃 𝐴𝑐𝑒) + 𝑃(𝑆𝑝𝑎𝑑𝑒 because we will
double count the Ace of Spades
 Of the 52 cards:
 4 are aces: 𝑃 𝐴𝑐𝑒 = 4Τ52
 13 are spades: 𝑃 𝐴𝑐𝑒 = 4Τ52
 1 is both an ace and a spade: 𝑃 𝐴𝑐𝑒, 𝑆𝑝𝑎𝑑𝑒 = 1Τ52
 4
𝑃 𝐴𝑐𝑒 𝑜𝑟 𝑆𝑝𝑎𝑑𝑒 = +⋯
52
 4
𝑃 𝐴𝑐𝑒 𝑜𝑟 𝑆𝑝𝑎𝑑𝑒 = +⋯
52
 4 13
𝑃 𝐴𝑐𝑒 𝑜𝑟 𝑆𝑝𝑎𝑑𝑒 = + +⋯
52 52
 4 13 1
𝑃 𝐴𝑐𝑒 𝑜𝑟 𝑆𝑝𝑎𝑑𝑒 = + − …
52 52 52
 4 13 1 𝟏𝟔
𝑃 𝐴𝑐𝑒 𝑜𝑟 𝑆𝑝𝑎𝑑𝑒 = + − =
52 52 52 𝟓𝟐
Addition Rule

 Consider two events A and B, not necessarily mutually exclusive.

 Let 𝑃(𝐴, 𝐵) be the joint probability that both A and B happen.

 Then 𝑃 𝐴 𝑜𝑟 𝐵 =
𝑃 𝐴 + 𝑃 𝐵 − 𝑃(𝐴, 𝐵)
 Combining probability rules
Hogwarts has 280 students: 70 each in Gryffindor, Ravenclaw, Hufflepuff,
and Slytherin. There are 40 first-year students, 10 in each house. A
student is randomly selected for the next Tri-Wizard Tournament.
What is 𝑃(𝑛𝑒𝑖𝑡ℎ𝑒𝑟 𝑌𝑒𝑎𝑟 1 𝑛𝑜𝑟 𝑅𝑎𝑣𝑒𝑛𝑐𝑙𝑎𝑤)?

𝑃(𝑌1 𝑜𝑟 𝑅) = 𝑃(𝑌1) + 𝑃(𝑅) − 𝑃(𝑌1, 𝑅)

40 70 10 100
= + − = ≈ 0.357
280 280 280 280

And so by the negation rule,

180
𝑃 𝑛𝑒𝑖𝑡ℎ𝑒𝑟 𝑌𝑒𝑎𝑟 1 𝑛𝑜𝑟 𝑅𝑎𝑣𝑒𝑛𝑐𝑙𝑎𝑤 = ≈ 0.643
280
Where do probabilities come from?

 Careful counting
 Data
 Subjective judgment
 Other probabilities
What questions are there?
Conditional Probability

 A conditional probability is the chance that one event (A)

happens given that some other event (B) has already happened.
 Our notation: 𝑃 𝐴 𝐵 is the “probability of A given B”
 Conditional probabilities reflect our uncertainty in light of partial
knowledge:
 𝑃 𝑟𝑎𝑖𝑛 𝑡ℎ𝑖𝑠 𝑎𝑓𝑡𝑒𝑟𝑛𝑜𝑜𝑛 𝑐𝑙𝑜𝑢𝑑𝑦 𝑡ℎ𝑖𝑠 𝑚𝑜𝑟𝑛𝑖𝑛𝑔
 𝑃 𝑈𝑇 𝑏𝑒𝑎𝑡𝑠 𝑂𝑈 𝑈𝑇 𝑎ℎ𝑒𝑎𝑑 𝑏𝑦 𝑎 𝑡𝑜𝑢𝑐ℎ𝑑𝑜𝑤𝑛 𝑎𝑡 ℎ𝑎𝑙𝑓𝑡𝑖𝑚𝑒)
 𝑃 𝑎𝑐𝑐𝑒𝑝𝑡𝑒𝑑 𝑡𝑜 𝑚𝑒𝑑𝑖𝑐𝑎𝑙 𝑠𝑐ℎ𝑜𝑜𝑙 𝑐𝑜𝑙𝑙𝑒𝑔𝑒 𝐺𝑃𝐴 > 3.6)
 …
Example: Recommender Systems

 Instagram: P follow @LeoMessi follow @Cristiano)

 Amazon: P buy organic dog food bought GPS dog collar
 Netflix: P watch 𝑇𝑖𝑛𝑘𝑒𝑟 𝑇𝑎𝑖𝑙𝑜𝑟 𝑆𝑜𝑙𝑑𝑖𝑒𝑟 𝑆𝑝𝑦 watch 𝑆ℎ𝑒𝑟𝑙𝑜𝑐𝑘)
Example: Recommender Systems

 Instagram: P follow @LeoMessi follow @Cristiano)

 Amazon: P buy organic dog food bought GPS dog collar
 Netflix: P watch 𝑇𝑖𝑛𝑘𝑒𝑟 𝑇𝑎𝑖𝑙𝑜𝑟 𝑆𝑜𝑙𝑑𝑖𝑒𝑟 𝑆𝑝𝑦 watch 𝑆ℎ𝑒𝑟𝑙𝑜𝑐𝑘)
Perhaps the single most important fact to
remember about conditional probabilities:

𝑃(𝐴|𝐵) ≠ 𝑃(𝐵|𝐴)
Perhaps the single most important fact to
remember about conditional probabilities:

𝑃(𝐴|𝐵) ≠ 𝑃(𝐵|𝐴)

A: You can dribble a basketball

B: You play in the NBA
𝑃 𝐶𝑎𝑛 𝑑𝑟𝑖𝑏𝑏𝑙𝑒 𝑃𝑙𝑎𝑦𝑠 𝑖𝑛 𝑁𝐵𝐴 = 1)
𝑃 𝑃𝑙𝑎𝑦𝑠 𝑖𝑛 𝑁𝐵𝐴 𝐶𝑎𝑛 𝑑𝑟𝑖𝑏𝑏𝑙𝑒 ≈ 0
Perhaps the single most important fact to
remember about conditional probabilities:

𝑃(𝐴|𝐵) ≠ 𝑃(𝐵|𝐴)
Moral of the story:
Always be specific about what’s on the left-hand side
and what’s on the right-hand side.
Where do probabilities come from?

 Careful counting
 Data
 Subjective judgment
 Other probabilities

It’s just the same for conditional probabilities!

 Example: Mammograms

P cancer = 15Τ200
P die, cancer = 3Τ200
P die|cancer = 3Τ15
In general, we can estimate 𝑃(𝐴|𝐵) as:

𝐹𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 𝑜𝑓 𝐴 𝑎𝑛𝑑 𝐵 𝑏𝑜𝑡ℎ ℎ𝑎𝑝𝑝𝑒𝑛𝑖𝑛𝑔

𝑃 𝐴𝐵 =
𝐹𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 𝑜𝑓 𝐵 ℎ𝑎𝑝𝑝𝑒𝑛𝑖𝑛𝑔
Multiplication Rule

 The probability of A given B is equal to the frequency of A and B

both happening divided by the frequency of B happen.

 The multiplication rule expresses this idea in general terms:

𝑃(𝐴, 𝐵)
𝑃 𝐴𝐵 =
𝑃(𝐵)
Multiplication Rule

 We can also use the alternate version below if we want to go in

reverse, from a conditional probability to a joint probability.

 This says the same thing with the terms re-arranged:

𝑃 𝐴, 𝐵 = 𝑃(𝐴 𝐵 ∗ 𝑃(𝐵)
Example: Mammograms

P cancer = 15Τ200
P die, cancer = 3Τ200
P die|cancer = 3Τ15
Using the multiplication rule:

𝑃(𝑑𝑖𝑒, 𝑐𝑎𝑛𝑐𝑒𝑟)
𝑃 𝑑𝑖𝑒 𝑐𝑎𝑛𝑐𝑒𝑟 =
𝑃(𝑐𝑎𝑛𝑐𝑒𝑟)
Example: Mammograms

P cancer = 15Τ200
P die, cancer = 3Τ200
P die|cancer = 3Τ15
Using the multiplication rule:

𝑃(𝑑𝑖𝑒, 𝑐𝑎𝑛𝑐𝑒𝑟)
𝑃 𝑑𝑖𝑒 𝑐𝑎𝑛𝑐𝑒𝑟 =
𝑃(𝑐𝑎𝑛𝑐𝑒𝑟)
3Τ200 3
= = ≈ 0.2
15Τ200 15
 Example: Mammograms

P cancer = 15Τ200
P die, cancer = 3Τ200
P die|cancer = 3Τ15
Or the re-arranged multiplication rule:

𝑃 𝑑𝑖𝑒, 𝑐𝑎𝑛𝑐𝑒𝑟 = 𝑃 𝑑𝑖𝑒 𝑐𝑎𝑛𝑐𝑒𝑟 ∗ 𝑃 𝑐𝑎𝑛𝑐𝑒𝑟

 Example: Mammograms

P cancer = 15Τ200
P die, cancer = 3Τ200
P die|cancer = 3Τ15
Or the re-arranged multiplication rule:

𝑃 𝑑𝑖𝑒, 𝑐𝑎𝑛𝑐𝑒𝑟 = 𝑃 𝑑𝑖𝑒 𝑐𝑎𝑛𝑐𝑒𝑟 ∗ 𝑃 𝑐𝑎𝑛𝑐𝑒𝑟

3 15 3
= ∗ = ≈ 0.015
15 200 200
Probability
and
Contingency
Tables
Pretend that you’re on the Data Science team at

 What we want to know (A): Maria likes Saving Private Ryan?

 What we already know (B): Maria likes Band of Brothers
 Key question: what is 𝑃(𝑨 | 𝑩)?
 Answer: go to the data and use the multiplication rule!

𝐹𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 𝑜𝑓 𝐴 𝑎𝑛𝑑 𝐵 𝑏𝑜𝑡ℎ ℎ𝑎𝑝𝑝𝑒𝑛𝑖𝑛𝑔

𝑃 𝐴𝐵 =
𝐹𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 𝑜𝑓 𝐵 ℎ𝑎𝑝𝑝𝑒𝑛𝑖𝑛𝑔
Probability and
Contingency Tables Liked Saving
Didn’t like it
Private Ryan

Liked Band of
Brothers
56 6

Didn’t like it
14 24
Probability and
Contingency Tables Liked Saving
Didn’t like it
Private Ryan

Liked Band of
Brothers
56 6

Didn’t like it
14 24

56
P Likes 𝑆𝑃𝑅 Likes 𝐵𝐵 = ≈ 0.9
56 + 6
 Some numbers

Network TV
 $130 million for one  $400 million
season of 10 episodes commissioning pilots
 7,000 individual period  113 shows
costumes
 13 make it to a
One $35,000 royal
second season


wedding dress (300 annual

Netflix subscriptions)
Band ACL Bonnaroo Coachella Lollapalooza Outside Lands

Jimmy Cliff 0 1 0 1 0

Pretty Lights 1 1 0 1 0

Lila Downs 0 0 0 0 1

Rebelution 1 1 0 1 0

Black Joe Lewis and the Honeybears 0 0 1 0 0

Explosions In The Sky 0 1 0 1 0

Brand New 0 0 0 1 0

Frank Turner 1 0 1 0 0

Local Natives 0 0 0 1 0

Nas & Damian Marley 0 0 1 0 0

(+ 1,228 more rows)

Probability and
Contingency Tables
Didn’t play Lolla Played Lolla

Didn’t play ACL

719 361
Played ACL
81 77

𝑃(𝑝𝑙𝑎𝑦𝑒𝑑 𝐴𝐶𝐿, 𝑝𝑙𝑎𝑦𝑒𝑑 𝐿𝑜𝑙𝑙𝑎)?

𝑃 𝑝𝑙𝑎𝑦𝑒𝑑 𝐴𝐶𝐿 𝑝𝑙𝑎𝑦𝑒𝑑 𝐿𝑜𝑙𝑙𝑎)?

Probability and
Contingency Tables
Didn’t play Lolla Played Lolla

Didn’t play ACL

719 361
Played ACL
81 77

77
𝑃 𝑝𝑙𝑎𝑦𝑒𝑑 𝐴𝐶𝐿, 𝑝𝑙𝑎𝑦𝑒𝑑 𝐿𝑜𝑙𝑙𝑎 = ≈ 0.062
(719 + 361 + 81 + 77)
Probability and
Contingency Tables
Didn’t play Lolla Played Lolla

Didn’t play ACL

719 361
Played ACL
81 77

77
𝑃 𝑝𝑙𝑎𝑦𝑒𝑑 𝐴𝐶𝐿 𝑝𝑙𝑎𝑦𝑒𝑑 𝐿𝑜𝑙𝑙𝑎) = ≈ 0.176
(361 + 77)
Let’s try this out together with the music festivals data!

Download aclfest files on Canvas: