Prof. Thistleton MAT 505 Introduction to Probability Lecture...
Transcript of Prof. Thistleton MAT 505 Introduction to Probability Lecture...
Prof. Thistleton MAT 505 Introduction to Probability Lecture 6
SUNY POLY Page 1
Sections from Text and MIT Video Lecture: Sections 1.2, 1.3, 1.4, and 4.1, 4.2
http://ocw.mit.edu/courses/electrical-engineering-and-computer-science/6-041-probabilistic-
systems-analysis-and-applied-probability-fall-2010/video-lectures/lecture-3-independence/
http://ocw.mit.edu/courses/electrical-engineering-and-computer-science/6-041-probabilistic-
systems-analysis-and-applied-probability-fall-2010/video-lectures/lecture-4-counting/
Topics from Syllabus: Counting, Random Variables: Discrete and Continuous, Probability mass
functions and probability densities
A (Ridiculously Brief) Introduction to Discrete Random Variables
Obviously we will deal with this topic in come complexity and depth, but for a start let’s see
what the simplest idea in probability might be. If I am going to give you the most meager bit of
information, I could tell you the result of a dichotomy, i.e. a Yes or a No. Can you think of
anything simpler?
We model this easiest of all situations with a “Bernoulli Random Variable”. For example,
consider a coin toss with a “fair” coin. Our sample space is given as
𝑆 = {𝐻, 𝑇}
You can immediately generalize to a polling situation where we are trying to determine the
proportion of “Lefties” amongst mathematicians. (I am thinking right brain/left brain, not
politics!) We poll for dichotomies all the time.
Prof. Thistleton MAT 505 Introduction to Probability Lecture 6
SUNY POLY Page 2
We already know from the Classical Notion of Probability that 𝑃(𝐻) =1
2= 𝑃(𝑇) . We can also
create a bookkeeping function that “counts” the number of Heads on this single toss. This may
be overkill in the current situation, but it will generalize nicely. Our function 𝑋 maps the sample
space 𝑆 into the real numbers ℝ.
𝑋: 𝑆 → ℝ
Definitions
Given a sample space S, we define a random variable X, to be a (measurable) function which
maps elements of the sample space, s ∈ S , onto real numbers, X(s).
Notice that a random variable induces probabilities on subsets of the real line.
Given a random variable 𝑋, define 𝑋 to be discrete if X may assume a finite or a countably
infinite number of values 𝑥 = 𝑋(𝑠). Also, define a function
𝑝(𝑥) ≡ 𝑃(𝑋 = 𝑥)
We call 𝑝(𝑥) the probability function of X or the probability mass function of X. We immediately
have the following (note the similarities with relative frequencies):
𝑝(𝑥) ≥ 0 ∀ 𝑥 ∈ ℝ
∑ 𝑝(𝑥)𝑥 = 1
Prof. Thistleton MAT 505 Introduction to Probability Lecture 6
SUNY POLY Page 3
We can make a table like this.
𝑠 ∈ 𝑆 𝑋(𝑠)
H 1
T 0
We can also make a table showing each of the possible real number outcomes and their
associated probabilities. Such a table represents a “probability mass function”, or pmf.
𝑘 𝑃(𝑋 = 𝑘) ≡ 𝑝(𝑘)
0 1
2
1 1
2
∑ 𝑝(𝑘)1
𝑘=0= 1
This is not bad so far. Just keep track of our notation- we let 𝑘 represent a possible real number
outcome from our experiment and call its probability mass function 𝑝(𝑘) . Can you build the
next case- tossing two fair coins? Here are our tables for 𝑛 ≡ number of tosses = 2.
Prof. Thistleton MAT 505 Introduction to Probability Lecture 6
SUNY POLY Page 4
𝑠 ∈ 𝑆 𝑋(𝑠) ≡ number of heads
HH 2
HT 1
TH 1
TT 0
𝑘 𝑃(𝑋 = 𝑘) ≡ 𝑝(𝑘)
0 1
4
1 1
4+
1
4=
1
2
2 1
4
∑ 𝑝(𝑘)2
𝑘=0= 1
It is not a bad practice to begin plotting our pmf’s.
Here is the code for the previous picture.
k=seq(0,2)
p=c(1/4, 1/2, 1/4)
plot(k, p,main="Probability Mass Function, n=2",xlim=c(-1,3),ylim=c(0,1.1))
lines(k,p, type="h")
Prof. Thistleton MAT 505 Introduction to Probability Lecture 6
SUNY POLY Page 5
Now, what happens when we consider larger values for 𝑛, ≡ the number of tosses. I can stand
to do the case for n=3 by hand, but this becomes tedious for larger values. A little theory can
help automate our results.
𝑠 ∈ 𝑆 𝑋(𝑠) ≡ number of heads
HHH 3
HHT 2
HTH 2
HTT 1
THH 2
THT 1
TTH 1
TTT 0
𝑘 𝑃(𝑋 = 𝑘) ≡ 𝑝(𝑘)
0 1
8
1 1
8+
1
8+
1
8=
3
8
2 1
8+
1
8+
1
8=
3
8
3 1
8
∑ 𝑝(𝑘)2
𝑘=0= 1
Prof. Thistleton MAT 505 Introduction to Probability Lecture 6
SUNY POLY Page 6
Also, we can’t help but notice that, if we watch the values of the numerators in our probabilities
a famous result becomes apparent. We will spend the rest of this lecture trying to understand the
two tables shown below, courtesy of Wikipedia:
Counting
We are now able to calculate probabilities in fairly simple situations. We would like to be able to
move towards more complex situations. For example, what is the probability of exactly 10 heads
when tossing a coin 20 times?
Here are a few situations where we need to learn how to “count”.
A set of 𝑛 simple coin tosses (also called a Binomial Random Variable) with a possibly
unfair or biased coin. In this case we try to describe the probability of obtaining 𝑟
successes on 𝑛 independent trials. This is modeled with a Binomial Random Variable.
Prof. Thistleton MAT 505 Introduction to Probability Lecture 6
SUNY POLY Page 7
A waiting time scenario where we keep tossing a (possibly unfair or biased) coin until
our first Heads/Success. (For example, when playing “Trouble”, we keep rolling a die
until our first "Six"). This is modeled by a Geometric Random Variable.
A waiting time scenario where we keep tossing a (possibly unfair or biased) coin until we
have obtained exactly 𝑟 Heads/Successes. We try to describe the probability that this will
take 𝑟 trials, 𝑟 + 1 trials, etc. This is modeled with a Negative Binomial Random
Variable.
Simple Counting Arguments
Remember back to when you were a kid playing with Mr. Potato Head.
Suppose you are putting together a Mr. Potato Head and you have in front of you 4 mouths, 3
noses and 2 sets of eyes. How many different faces can you make?
We can think about the problem this way: We can choose a mouth in any of 4 different ways:
Start
Mouth 1
Mouth 2
Mouth 3
Mouth 4
Prof. Thistleton MAT 505 Introduction to Probability Lecture 6
SUNY POLY Page 8
Now we will select a nose. We can select a different nose with any of the 4 mouths:
It’s already looking pretty busy. The point is that with 4 Mouths and 3 Noses we can put together
4 times 3 = 12 different faces. If we include the eyes we can do 4 ∙ 3 ∙ 2 = 24 different faces.
Multistage Process: Suppose a process will be performed in N stages. For example, N = 4
when we toss a fair coin 4 times. Each of these N = 4 stages can be done in two ways (HEADS
or TAILS) so there are 2 ∙ 2 ∙ 2 ∙ 2 = 16 ways to toss a coin (fair or biased) 4 times.
In the Mr. Potato Head example, there are 3 stages: Select a mouth, select a nose, select some
eyes. The first stage may be done in 4 ways, the second in 3 ways and the last in 2 ways.
Key Example: Here’s an example that is solved every day many times all around the world.
We’ll present it in a simplified way. Suppose you wish to visit 5 cities: Albany, Utica,
Binghamton, Rochester, and Syracuse. You will visit each one once and only once. How many
ways can you do this?
By simply using the multistage process approach we see that we can choose the
first city in 5 ways. Say that we start from Utica. Once that’s chosen there are 4
Start
Mouth 1
Nose 1
Nose 2
Nose 3
Mouth 2
Nose 1
Nose 2
Nose 3
Mouth 3
Nose 1
Nose 2
Nose 3
Mouth 4
Nose 1
Nose 2
Nose 3
Prof. Thistleton MAT 505 Introduction to Probability Lecture 6
SUNY POLY Page 9
ways to select from the remaining cities (Albany, Binghamton, Rochester, and
Syracuse). If you were only visiting 2 out of 5 cities there would be 5 ∙ 4 = 20 .
We’re visiting all 5, so if we choose, say Utica for the first and Rochester for the
second then selecting the third from Albany, Binghamton, and Syracuse means
there are 5 ∙ 4 ∙ 3 = 60 ways to choose 3 cities. The fourth comes from the
remaining 2, then the last is determined for us. All told: 5 ∙ 4 ∙ 3 ∙ 2 ∙ 1 =
120 ways to visit 5 cities.
Let’s push this a little further. Suppose you want to visit each state capitol once and only once.
There are 50 states, so there are 50 ways to choose the first. If we were only visiting 2 of them
there would be 50 ∙ 49 = 2450 ways to plan our trip. For 3 cities we would have 50 ∙ 49 ∙ 48 =
117600. You should be starting to get a little worried. Build a table:
Number of Cities Number of ways to plan trip
1 50
2 50∙49 = 2450
3 50∙49∙48 = 117600
4 50∙49∙48∙47 = 5527200
5 50∙49∙48∙47∙46 = 254251200
6 50∙49∙48∙47∙46∙45 11441304000
By the time we get to 50 cities there are 3.041409320171338e+064 ways. This number is so
large as to be incomprehensible. I can barely understand one million, but it helps to think about a
screen door.
Prof. Thistleton MAT 505 Introduction to Probability Lecture 6
SUNY POLY Page 10
Some Notation: Factorials When we multiply numbers like “3 times 2 times 1” we will
usually write 3 ∙ 2 ∙ 1 = 3! which is read as “3 factorial”. From the above we have
50 ! ≈ 3.041409320171338e + 064
At this point we can probably use some refreshments.
Suppose you have 3 candy bars in front of you, Snickers, a Three Musketeers, and a Baby
Ruth. Call them S, T, and B. In how many ways can you arrange these objects in order?
3! = 3 ∙ 2 ∙ 1 = 6
Here they are: STB, SBT, TSB, TBS, BST, BTS.
(Just as a side note: I taught High School some time ago in Cameroon. Some students needed
help in combinatorics or counting arguments, (which I had never looked at myself) and I
foolishly agreed to help them. I wasn’t able to really “click” with the material until I physically
placed objects in front of me on the table top and started manipulating them. You are strongly
encouraged to do the same in order to make this all less abstract.)
Suppose now you add a Hershey bar, call it H. How many arrangements are possible now?
4! = 4 ∙ 3 ∙ 2 ∙ 1 = 24
Write them down on a piece of paper!
How many when you add a Reese’s, R?
5! = 5 ∙ 4 ∙ 3 ∙ 2 ∙ 1 = 120
You can be forgiven if you don’t write these down, but make sure you understand where the
numbers are coming from. Stay with the 5 bars. Suppose you are going to eat three of them, and
it matters to you in what order you eat them. In how many ways can this be done?
Prof. Thistleton MAT 505 Introduction to Probability Lecture 6
SUNY POLY Page 11
5 ∙ 4 ∙ 3 = 60
Did you notice that we started out in the form of 5! But lost the last two numbers? We can write:
5 ∙ 4 ∙ 3 =5 ∙ 4 ∙ 3 ∙ 2 ∙ 1
2 ∙ 1=
5!
2!= 60
Permutations:
The number of possible sequences (ordered arrangements) of n objects is
𝑛! = 𝑛 · (𝑛 − 1) · (𝑛 − 2) · . . . 3 · 2 · 1
If the arrangement is to contain only r of these n objects, then the number of ways to form
sequences is called the permutation of n things taken r at a time and we have
𝑃𝑛,𝑟 ≡ 𝑛!
(𝑛 − 𝑟)!
As another example, suppose you are dealing out 5 cards from a normal 52 card deck. If the
order in which the cards come to you is important (like when you bet on each card before
moving on), then the number of possible “hands” is
52 ∙ 51 ∙ 50 ∙ 49 ∙ 48 = 𝑃52,5 =52!
(52−5)!=
52!
(47)!= 311875200 .
That’s a lot!!!
Combinations
This is the last idea from Chapter 4. Consider again the 5 candy bars: S, T, B, H, and R.
Sometimes we don’t care about the order in which we select items. How many ways may you
select 3 of these candy bars if the order of the selection is unimportant to you? That is, STH is
the same TSH?
Prof. Thistleton MAT 505 Introduction to Probability Lecture 6
SUNY POLY Page 12
Let’s take this from the top. If you are very patient, you can write out all
𝑃5,3 ≡ 5!
(5 − 3)!=
120
2= 60
ways to select a candy bar if order is important and then start crossing off the sequences that are
really equivalent, or just rearrangements.
Luckily, I am this patient.
STH SHT TSH THS HST HTS
STR SRT TSR TRS RST RTS
STB SBT TSB TBS BST BTS
SHR SRH HSR HRS RSH RHS
SHB SBH HSB HBS BSH BHS
SRB SBR RSB RBS BSR BRS
THR TRH HTR HRT RTH RHT
THB TBH HTB HBT BTH BHT
TRB TBR RTB RBT BTR BRT
HRB HBR RHB RBH BHR BRH
BCD BDC CBD CDB DBC DCB
Prof. Thistleton MAT 505 Introduction to Probability Lecture 6
SUNY POLY Page 13
If you have less patience, reason as follows: Within the 60 possible sequences, how many are
really just TSH rearranged? Evidently that would be 3 ∙ 2 ∙ 1 = 6 If we “divide these out” we
come up with
𝑃5,3
3!≡
5!
(5 − 3)! 3!=
120
2 ∙ 6= 10
These we should all be happy to list by hand: S, T, B, H, R select 3, order unimportant
STB SBH SHR TBH THR BHR
STH SBR TBR
STR
More formally, the number of possible unordered groups which contain r out of n objects is
simply the number of ordered groups, i.e. the number of possible permutations, 𝑃𝑛,𝑟 divided by
the number of ways to arrange r objects (i.e. we eliminate the redundant groups). Therefore
𝐶𝑛,𝑟 ≡ 𝑃𝑛,𝑟
𝑟! =
𝑛!
(𝑛 − 𝑟)! 𝑟!
Another Key Example Suppose you toss a biased coin 6 times. We know that there are 2 ∙ 2 ∙
2 ∙ 2 ∙ 2 ∙ 2 = 26 = 64 possible sequences:
Prof. Thistleton MAT 505 Introduction to Probability Lecture 6
SUNY POLY Page 14
HHHHHH HHTHHH HTHHHH HTTHHH THHHHH THTHHH TTHHHH TTTHHH
HHHHHT HHTHHT HTHHHT HTTHHT THHHHT THTHHT TTHHHT TTTHHT
HHHHTH HHTHTH HTHHTH HTTHTH THHHTH THTHTH TTHHTH TTTHTH
HHHHTT HHTHTT HTHHTT HTTHTT THHHTT THTHTT TTHHTT TTTHTT
HHHTHH HHTTHH HTHTHH HTTTHH THHTHH THTTHH TTHTHH TTTTHH
HHHTHT HHTTHT HTHTHT HTTTHT THHTHT THTTHT TTHTHT TTTTHT
HHHTTH HHHTTTH HTHTTH HTTTTH THHTTH THHTTTH TTHTTH TTTTTH
HHHTTT HHTTTT HTHTTT HTTTTT THHTTT THTTTT TTHTTT TTTTTT
o How many of these have exactly 0 HEADS:
That’s easy, just TTTTTT
o How many of these have exactly 1 HEADS:
That’s not too bad, just look through the list to see HTTTTT, THTTTT, TTHTTT,
TTTHTT, TTTTHT, TTTTTH
o How many of these have exactly 2 HEADS:
This is starting to become work. Here’s how we think about it: envision 6 places,
like 6 cities, candy bars, etc. Out of these 6, select 2 of them. This is a
combination because selecting the third slot for a H and then the fifth gives the
same outcome TTHTHT as selecting the fifth and then the third. Evidently we
should find
𝐶6,2 ≡ 𝑃𝑛,𝑟
𝑟! =
6!
(6 − 2)! 2!=
720
24 ∙ 2= 15
Prof. Thistleton MAT 505 Introduction to Probability Lecture 6
SUNY POLY Page 15
Of these, take a second and with a highlighter indicate these 15 slots.
o How many of these have exactly 3 HEADS:
𝐶6,2 ≡ 𝑃𝑛,𝑟
𝑟! =
6!
(6 − 3)! 3!=
720
6 ∙ 6= 20
o You should do the same for 4, 5, and 6 HEADS.
o
More Examples
How many (5 card) poker hands are possible if the order of the cards is unimportant?
o How many of these are full houses?
o How many are straight flushes, etc?
In how many distinct ways can you arrange the letters: a, b, c, d?
In how many distinct ways can you arrange the letters: b, o, o, k?
The letters: m, i, s, s, i, s, s, i, p, p, i?
There are 10 math majors, 15 engineering majors, and 5 students from other majors in a
probability class. In how many ways can you select two students from this class in such a
way that they are from different majors?
In how many ways can a set of 𝑛 elements be arranged if there are 𝑛1 objects of one type, 𝑛2
objects of another type, and so on till 𝑛𝑘 objects of the 𝑘𝑡ℎ type, where
∑ 𝑛𝑖 𝑘
𝑖=1= 𝑛
Prof. Thistleton MAT 505 Introduction to Probability Lecture 6
SUNY POLY Page 16
How many different binary numbers of length 7 are possible if 3 of the digits are 1’s?
In how many ways can a group of 8 people be divided into working groups of 2 people each?
In how many ways can a set of 𝑛 elements be partitioned into 𝑘 subsets with 𝑛1,
𝑛2, … , 𝑛𝑘 elements in each subset?
An excellent reference for combinatorics is the text Applied Combinatorics by Alan Tucker.