LECTURE 1: Probability models and axioms

Readings:• Sample 1.1,

• Sections
Sample space
1.2
space
Probability laws
• Probability
• laws

– Axioms
– Axioms Lecture outline
– Some
Some properties
– Properties
properties
that follow from the axioms
Examples
• Examples
•
Discrete
– Discrete
–
Continuous
– Continuous
–
Discussion
• Discussion
•
Countable additivity
– Countable
– additivity
Mathematical subtleties
– Mathematical
– subtleties
Interpretations of
••• Interpretations
Interpretations of probabilities
probabilities
 Sample space

• List (set) of possible outcomes, Ω

• List must be:
– Mutually exclusive
– Collectively exhaustive
Sample space
– At the “right” granularity

ssible outcomes
• Two steps:
– Describe possible outcomes
sive
– Describe beliefs about likelihood of outcomes
haustive
granularity
 Sample space
Sample space
ssible outcomes Sample space
• List (set) of possible outcomes, Ω
• List (set) of possible outcomes, Ω
• List must be:
sive
• – List must be:
Mutually exclusive
haustive
granularity Mutually exclusive
– Collectively exhaustive
Collectively
– At exhaustive
the “right” granularity
– At the “right” granularity
 Sample space: discrete/finite example
Sample space: discrete/finite example
Sample space: discrete/finite example
• Two rolls of aSample tetrahedral
space: diediscrete/finite example
• Two rolls of a tetrahedral die
rolls of a tetrahedral die
– space• Two rolls
Sample of a vs.
space tetrahedral –dieSample
sequential description
space vs. sequential description
mple vs. sequential description
Die roll
– Sample space vs. sequential example
description 1,1
1,1 1,1 1 1,2
1 1,2
1,3
1,1 1 1,2 1,3
1 1,2 1,4
1,4 4 1,3 1,3
4 4 2
2 1,4 1,4
44 Y = Second 3
Y = Second 3 2 2
roll
3 3
Y = roll
Second
2 Y = Second 33 3 2
roll Y = Second
roll
3 1
1
2 roll 2
3
1 2 3 4 12 1 2 3 4
4
4
1 X = First roll 3
X = First roll 4,4
1 1 2 44,4
4
X = First roll
4,4
ntinuous sample space: 1 2 3 •1 A
4 continuous
2 3 4 sample space:
4
A continuous
) such that 0 ≤ x, y ≤ 1
• sample space: (x, y) such that 0 ≤ x, y ≤ 1
X = First roll X = First roll 4,4
(x, y) such
y that 0 ≤ x, y ≤ 1 y
• A continuous
• Let B be1thesample space:
y Y)=2
event: min(X, 1
(x, •y)Let
such
M =that 0 Y≤) x, y1≤ 1
max(X,

• P(M = 1 | B) = y
1 x1 1 x
mple space: continuous example
mple space: Sample space: continuous example
Sample space: sample
• A continuous continuous
space:example
ntinuous sample space:
Sample space: continuous example
• (x, y) such that 0 ≤ x, y ≤ 1
0 ≤ x, y ≤ 1
inuous sample space:
such that 0space: 1
≤ x, y ≤continuous
Sample example
y
us sample space:
uch that 0 ≤ x, y ≤ 1
y
1y
that 0 ≤ x, y ≤ 1
1

1
Sample space: continuous
Sample example example
space: continuous
• A continuous sample x sample space:
1 space:
• A continuous

1 x
y ≤y10 ≤ x, y ≤ 1
• (x, y) such that
• (x, y) 0such
y
≤ x, that
y11
xy
1 1 1

1 x
 Probability axioms
Probability axioms
et of the•sample space
Event: a subset of the sample space
assigned –
to Probability
events is assignedProbability
to events axioms
Probability axioms
Probability
Probability axioms
axioms
• •Event:
Event: a subset
a subset
subset of of
thethe
sample
sample space
space
• Event: a subset of the sample space
•
Axioms:Event: a of the sample space
– –Probability
Probability is is
assigned
assignedto toevents
events
1. P– Probability
–(A) ≥0
Probability is
is assigned
assigned toto events
events
• •Axioms:
Axioms:
1 • Axioms:
Axioms: = 1
2. •P(universe)
1. P –(A) ≥0
Nonnegativity: (A) ≥
P(A) ≥ 00
–
– Nonnegativity:
Nonnegativity: P
P (A) ≥ 0
3. If A ∩ B = Ø,
= P(A)2. P–(universe)
+ Pthen
(B) =1
Normalization:
Normalization: P(Ω) = 11
P (A ∪ B)
– Normalization: = P(A) P(Ω)
P +P
(Ω) =
=(B)1
3. If– A(Finite)
∩ B = Ø, additivity:
additivity:
– (Finite)
thenIf P (A ∪ additivity:
B) = P (A) +(to
P be strengthened later)
(B)
A∩
If A ∩BB=
Ø, then
= Ø,
then P
Ø, then P (A ∪
P(A
(A ∪
∪ B) =P
B) =
B) = P (A)+
P(A) +P
(A) +
(B)
P(B)
P (B)

k }) = P({s
• 1}) + 1· ,· s· 2+
P({s . .({s
, .P }) = P({s1}) + · · · + P({s })
, skk}) k

= P(s1) + · · · + P(sk ) = P(s1) + · · · + P(sk )

• P({s 1 , s2
({s , ,s.22., .. ,. s. k, }) ==P({s ++
1 })})
({s · · ·· + P({s
····+ k })})
({s
•• P 1 skk})
}) = P
P ({s 1 })
P({s1, s2, . . . , sk }) = P({s1}) + · · · + P({skkk})
1 1 + · + P
P ({s })
==P(s
P 1
(s)+ ) · · ·· +
+ · · +P(sPk ) )
(s
= P(s11) + · · · + P(skkk))
= P (s 1 ) + · · · + P (s
ds strengthening
– Axiom 3 needs strengthening
 • Axioms
• Axioms
• Consequences
• Axioms
• Consequences
• Consequences
• Axioms For disjoint sets:
Probability axioms
Probability axioms • Axioms For disjoint sets:
• Consequences• Axioms
For disjoint sets:
• Consequences
a subset
subset of of Probability
the sample
sample axioms
space • Consequences
P(A ∪ B) = P(A) + P(B)
a Some the simple space consequences of the P(A axioms
∪ B) = P(A) + P(B)
For disjoint sets: • Axioms
ility
Pility
bset ,issof
({s1is assigned
, .the
2assigned to
=P
k }) to
. . , ssample events
({s
space
events 1}) + · · · + P({sP k(A ∪ B) = P
}) disjoint
For (A) + P(B)
sets:
• Axioms For disjoint
• Consequences sets: Some simple consequences of the ax
is assigned to events Some simple consequences of the a
= P (s
P (A
1 ) +∪ ·B)
· · +=PP(s (A)
k ) + P (B)
gativity: • Consequences
P (A) ≥ 0 P(A ∪•B)P= (A) P(A)≤Some1 + Psimple (B) consequences of the axioms
ativity: •P(A) ≥
Axioms 0 P(A ∪For • P (A) ≤
= P(A) sets:
B) disjoint 1 + P(B)
ization:
vity: (A)
PFor (Ω)
Pdisjoint
≥ 0= =1 Some• P(A) • ≤
simple P 1 consequences
(Ø) =0 of the axioms
ization: P (Ω)
• Consequences 1 sets: • P (Ø) = Some 0 simple consequences of the axioms
)ion:
additivity:
P (Ω) = (to
1 P(A) ≤ 1
be• strengthened • P(Ø)
later) P• (A =
A, ∪0B,B)CSome P(A)simple
=disjoint: +P P (B)
(A consequences
∪ B ∪ C) = P (A) of+the
P axioms
(B) + P(C
additivity: (to be strengthened later) • P(A) • ≤ A,1B, C disjoint: P(A ∪ B ∪ C) = P(A) + P(B) + P(
B = Ø,Pthen (AFor ∪P P (A=
B) ∪P
disjoint B) (A) =+ P(A)
events: P(B) + P(B)• P(A) (A) + ≤andP1 (A similarly
c ) = 1 for k disjoint events
= Ø, then (A ∪ B)
ditivity: (to be strengthened later)• =P P (A)
(Ø) = +0 P (B) and similarly for k disjoint events
• P(Ø) = 0 Some simple consequences of the ax
Ø, then P(A ∪ B) =• PA, (A) +CPdisjoint:
(B) • A, (Ø)
P(AB, • ∪CP= ({s
0 C)1,,ss2
disjoint: = , .P.P.(A)sk })
,(A ∪
+ B= ∪P
(B) ({s
C) + =1}) P +
(A)
(C) · ·+ · +PP ({s+
(B) k })P(C)
Some B, simple P
consequences• B
P ({s
∪ 1 of 2 , .
the. . , s })
axioms
k =
P P ({s 1P}) + · · · + P ({s k })
P(A ∪ B) = P(A) + P(B) • and A, • P(A)
B,similarly
C ≤1
disjoint: for kP (A ∪ B ∪events
Pdisjoint C) = P(A) + P(B) + P(C)
and similarly for • kA, disjoint
B, C events
disjoint: (A ∪ B ∪ C) = P·(A) +PP (B) + P(C)
• P(A) ≤ 1 and similarly for k disjoint= P (s
events
1 ) + · · + (s k )
P (Ø)
• similarly = 0})k=disjoint = P(s1) + · · · + P(sk )
• PSome , . . . , sk•})
({s1, s2simple and
P=({sP ,
({s s ,
}). . .
+ , sfor
· · · + P P
({s({s 1 })events
+ · · · + P({sk })
1
consequences 2
1 kof the k })
axioms
• P(Ø) = 0 • P({s1, s2, . . . , sk }) = P({s1}) + · · · + P({sk })
• P({s • A, B, C disjoint: = P({s P(A}) ∪ B ·∪· ·C)+= P(A) + P(B) + P(C
• P(A) ≤ 1 1 , s2 , . . . , sk }) = P (s )1+ + · · · + P P({s
(s ) k })
= P(s and1) + · · · + P(s k )k 1disjoint events k
• A, B, C disjoint: P(A ∪ B ∪ C) = P(A) +similarly
P(B) += for
P (C)
P(s1) + · · · + P(sk )
•and P(Ø) = 0 for k disjoint events
similarly = P(s1) + · · · + P(sk )
• P({s1, s2, . . . , sk }) = P({s1}) + · · · + P({sk })
P(A) + P(Ac) = 1 – Axiom 3 needs strengthening
• •P({s 1 , s2 , . . . , sk }) = P({s1 }) + · · · +– P({s Axiomk }) 3 needs strengthening
– Axiom – Do3 needs
weird sets = P(s
strengthening
have 1 ) + · · · + P(sk )
probabilities?
• A, B, C disjoint: P(A ∪ B ∪ C) = – P(A) Do+weird P(B)sets + Phave(C) probabilities?
= P(s1) + ·–· · + Do P(s weird
k ) sets have probabilities?
and similarly – Axiom 3 needsevents
for k disjoint strengthening
– Axiom 3 needs strengthening
• P({s1, s2, .–. . ,Do sk })weird
= P({s sets}) have
–
1 – +Do · · probabilities?
Axiom ·+
weirdP3({s needs strengthening
k }) have probabilities?
sets
Axiom 3 needs strengthening
– Do weird sets have probabilities?
 Probability axioms
Probability axioms

a subset
subset of of Probability
the sample axioms
space
a the sample space
Some simple consequences of the axioms
ility
Pility
bset ,issof
({s1is assigned
, .the to
=P
k }) to
. . , ssample
2assigned events
({s 1}) + · · · + P({sk })
space
events
• Axioms
is assigned to events = P(s1) + · · · + P(sk )
gativity: • Consequences
P (A) ≥ 0
ativity: •P(A) ≥
Axioms 0
ization:
vity: (A)
PFor (Ω)
Pdisjoint
≥ 0= =1
ization: •P(Ω) 1 sets:
Consequences
)ion:
additivity:
P(Ω) = (to (to
1 be be strengthened later)
additivity: strengthened later)
B = Ø,Pthen (A (A= = P(A)
(A) + P(B)
=+ P(B)
For ∪P PB) ∪PB)
disjoint events:
= Ø, then
ditivity: (to (A P(A)
be∪strengthened
B) +later)
P(B)
Ø, then P(A ∪ B) = P(A) + P(B)
P(A ∪ B) = P Some
(A) +simple
P(B) consequences of the axioms
• P(A) ≤ 1
Some simple consequences of the axioms
• P(Ø) = 0
• P(A) ≤ 1
• A, B, C disjoint: P(A ∪ B ∪ C) = P(A) + P(B) + P(C)
P(Ø)
•and = 0 for k disjoint events
similarly
P(A)
• •P({s + P(Ac) = 1
1 , s2 , . . . , sk }) = P({s1 }) + · · · + P({sk })
• A, B, C disjoint: P(A ∪ B ∪ C) = P(A) + P(B) + P(C)
= P(s1) + · · · + P(s )
and similarly for k disjoint events k
• P({s1, s2, . . . , sk }) = P({s1}) + · · · + P({sk })
Axiom 3 needs strengthening
 P(A ∪ B) = P(A) + P(B)

• Axioms Some simple consequences of the axioms

• Consequences
• P(A) ≤ 1
•ForP(Ø) = 0sets:
disjoint
Some simple consequences of the axioms
c =1
. ,(A)
P({s1, s2•, . .P sk })+=PP(A
({s)1}) + · · · + P({sk })
P(A ∪ B) = P(A) + P(B)
• A, B, C disjoint: P(A ∪ B ∪ C) = P(A) + P(B) + P(C)
= P(s1) + · · · + P(sk )
and similarly for k disjoint events
Some simple consequences of the axioms
• P({s1, s2, . . . , sk }) = P({s1}) + · · · + P({sk })
• P(A) ≤ 1
• P(Ø) = 0 = P(s1) + · · · + P(sk )

• P(A) + P(Ac) = 1
• A, B, C disjoint: P(A ∪ B ∪ C) = P(A) + P(B) + P(C)
– Axiom 3 needs strengthening
and similarly for k disjoint events
– Do weird sets have probabilities?
• P({s1, s2, . . . , sk }) = P({s1}) + · · · + P({sk })

= P(s1) + · · · + P(sk )

– Axiom 3 needs strengthening

– Do weird sets have probabilities?

Axiom 3 needs strengthening

 More consequences of the axioms
More consequences of the axioms
, then P(A) ≤ P(B)
• If A ⊂ B, then P(A) ≤ P(B)
) = P(A) + P(B) − P(A ∩ B)
• P(A ∪ B) = P(A) + P(B) − P(A ∩ B)
) ≤ P(A) + P(B)
• P(A ∪ B) ≤ P(A) + P(B)
∪ C) = P(A) + P(Ac ∩ B) + P(Ac ∩ B c ∩ C)
More consequences of the axioms
• P(A ∪ B ∪ C) = P(A) + P(Ac ∩ B) + P(Ac ∩ B c ∩ C)
• If A ⊂ B, then P(A) ≤ P(B)
• P(A ∪ B) = P(A) + P(B) − P(A ∩ B)
• P(A ∪ B) ≤ P(A) + P(B)
More consequences of the axioms
• P(A ∪ B ∪ C) = P(A) + P(Ac ∩ B) + P(Ac ∩ B c ∩ C)
• If A ⊂ B, then P(A) ≤ P(B)
• P(A ∪ B) = P(A) + P(B) − P(A ∩ B)
• P(A ∪ B) ≤ P(A) + P(B)
• P(A ∪ B ∪ C) = P(A) + P(Ac ∩ B) + P(Ac ∩ B c ∩ C)
 More consequences of the axioms
• If A ⊂ B, then P(A) ≤ P(B)
• P(A ∪ B) = P(A) + P(B) − P(A ∩ B)
More consequences of the axioms
• P(A ∪ B) ≤ P(A) + P(B)
, then P(A) ≤ P(B)
• P(A ∪ B ∪ C) = P(A) + P(Ac ∩ B) + P(Ac ∩ B c ∩ C)
) = P(A) + P(B) − P(A ∩ B)
) ≤ P(A) + P(B)
∪ C) = P(A) + P(Ac ∩ B) + P(Ac ∩ B c ∩ C)
 Example
4
4
3
Y = Second
roll 3
Y = Second
2
roll
1 2

11 2 3 4
Probability calculation: discrete/finite example X =1First roll
2 3 4
Sample space: discrete/finite example
X = First roll
Exampledie
• Two rolls of a tetrahedral • Let
Letevery possible
Z = min(X, Y )outcome have probability 1/16
– Sample space vs. 4sequential • P•(XLet
description
Die roll example
every
= 1) = possible outcome have probability 1/16
• P(Z = 1) =
3 • P(X = 1) 1,1
=
Y = Second
1 1,2
roll • P(Z = 2) =
1,3
2
4 1,4
1 4 Let Z =2 min(X, Y )
• P(Z = 3) =
Y = Second 31
Y = Second 3 •LetP(Z
Z ==min(X,
1) = Y )
2 3 4
roll
roll 2 X = First roll • P(Z = 3 4) =

2 1 •• P P(Z
(Z =
= 1)
2) =
=
t every possible outcome have 1 2 3 4
1 4
bability 1/16 X = First roll P(Z
•• P (Z =
= 2)
3) =
=
4,4

1 2 3 4
X = 1)•=A continuous sample space: P(Z
•• P (Z =
= 3)
4) =
=
(x, y) such that X
0 =≤First
x, yroll
≤1
• P(Z = 4) =
y
• Let B be the event: min(X, Y ) = 2
1
• Let M = max(X, Y )

• P(M = 1 | B) =
 Discrete uniform law

− Assume Ω consists
Discrete of n equally
uniform law likely element
Discrete uniform law
− Assume A consists of m elements
Discrete uniform law
Discrete uniform law
number of elements of A
e points be equally likely Then−: Assume
P(A) = Ωnumber
consistsofofelements
n equallyoflikel
Discrete uniform law − Assume Ω consists Ω
of
− Assume A consists of m elements
− Assume Ω consists of n equally likely elements
Assume Aofconsists of of elements
− Assume A consists of
• Let all −
sample points
number be equally
elements A • Just count. . .
klikely
P(A) = number of eleme
total number of sample points number of Then =
elements of A : P(A) =
k num
• Then, Then : P (A) =
number of elements of Ω n
Then : P(A)
number =
of eleme
number of elements of A
num
P(A) =
• Just count. . . total number of sample points
• Just count. . . 1
•prob = count. . .
Just
n
• Just count. . .
1
prob =
Continuous uniform law n

m” numbers in [0, 1].

y Continuous uniform law
1
• Two “random” numbers in [0, 1].
y
1
1 x

: Probability = Area
 Probabilitynumbers
• Two “random” calculation:
in [0,continuous
1]. example
Probability calculation: continuous exam
y
• Two “random” numbers in [0, 1].
• Two “random” numbers
y in [0,
1 1].
mple space: continuous example 1 y
mple space: Sample space: continuous example 1
Probability calculation: continuous example
• A continuous sample space:
1 x
Two “random” numbers in [0, 1]. 1 x
Sample space: continuous example
≤ x, y ≤ 1 (x, y) such that
• space:
uous sample
y 0 ≤ x, y ≤ 1 • Uniform probability law: Probability =x Area
1
Sample space: continuous example • Uniform probabilty law: Probability = Area
1
y
s sample space:
h that 0 ≤ x, y ≤ 1
�
x + y ≤ 1/2}
�
� law:=Probability = Area
P {(x, y) |probability
• �Uniform
y P {(x, y) | x + y ≤ 1/2} =
1y
hat 0 ≤ x, y ≤ 1
1 � � � �
1 x P� {(x, y) | 0.3)}
P {0.5, x �+ y ≤=1/2} =
P {0.5, 0.3)} =
1
� �
Uniform law: Probability = Area
Sample
Sample space: P {(0.5, 0.3)} =
space: continuous
continuous example example
1 x
• A continuous sample space:
• A continuous sample space:

1 x
• (x, y)0such that
≤ 10 ≤ x, y ≤ 1
• (x, y) such that ≤ x, y
y1
xy
1 y1
1

1 x 1 x

1 x
 Probability calculation steps
Probability calculation steps
he sample space
• Specify the sample space
probability law
• Specify a probability law
n event of interest
• Identify an event of interest
...
• Calculate...
ability calculation: discrete but infinite sample space
Probability calculation: discrete but infinite sample space
pace: {1, 2, . . .}
• Sample
1 space: {1, 2, . . .}
given P(n) = n , n = 1, 2, . . .
2 1
– We are given P(n) = n , n = 1, 2, . . .
me is even) = 2
• P(outcome is even) =
p
p
1/2
1/2
1/4
1/8 1/4
1/16
 Probability calculation: discrete
Probability discrete but
but infinite
infinite sample
sample space
space

Sample space:
•• Sample space: {1, 2, . . .}
–– We are given P(n) = 2−n, n
We are n== 1,
1,2,
2,......
Probability calculation: ••discrete Find P
Find (outcome
P(outcome
but infiniteis even)
sample space
Probability calculation: discrete but infinite sample space
• Sample Probability
space: {1, 2, calculation:
. . .} discrete but pinfinite sample space
• Sample space: {1, 2,−n . . .}
– We Sample
• are givenspace:
P(n) ={1, 2 2, ,. .n.}= 1, 2, . . .
– We are given P(n) = 2−n 1 , n = 1, 2, . . . 1/2
1/2
• Find –P(outcome is even)
We are given P(n) = n , n = 1, 2, . . .
• Find P(outcome
Probability is even)2 discrete but infinite sample
calculation: 1/4 space 1/4
• Find P(outcome
p is even)
…..
…..
1/8
1/8
• Sample space: {1,p 2, . . .} 1/16
1/16

p1/2 1
– We are given P(n) = 1/2 , n = 1, 2, . . . 1 22 33 44
2n
1/4 1/2
• P(outcome is even) = 1/4
Solution:
•• Solution:
…..
1/8
1/16
p
1/4 1/8
…..
1/16
P({2, 4, 6, . . .}) =
= PP(2)
(2) +
+P P(4)
(4)+ +······
1 2
1
1/2
3
2 3
41/8
4 …..
1/16
=
11
= 22 +
22
11
+ 44 +
22
11
+ 66 +
22
+······=
=
11
33
1 2 3 4
• Solution:
• Solution: •• Axiom Axiom needed:
1/4 needed:
P({2, 4, 6, . . .}) = If P(2)
If A A22+ ,, .P. .(4)
are+disjoint
··· events, then:
events, then:
…..
• Solution: A 1,, A 1/8
P({2, 4, 6, . . .}) 11= P1(2) +1/16 1 P (4) + ··1 ·
P({2, 4, 6,= . . .})
2=2=1
+ 4 + 16 + ·1· · =
2P+ (2)2+ P(AP(4)1 ∪+A2·3·∪· · · ·1
· )) =
=P (A11))+
P(A +P (A22))+
P(A +· ·· ·· ·
1 2 321 4 1+ 61+ · · · =
2 2 2 31
= + + + · · · =
• Axiom needed: 22 24 26 3
• Axiom
If •A1Solution: needed:
, A2, . . . are disjoint events, then:
• If Axiom
A1, A2,needed:
. . . are disjoint events, then:
If A1, A2P , .(A ({2,
. .P1are 4,∪6,· ·. ·. ).})
∪ Adisjoint
2 =P =(AP
events, )(2)
1then:++ P(AP(4)2) ++· · · ·
 Countable additivity axiom
Countable additivity axiom
Countable additivity axiom
ens the finite additivity axiom
• Strengthens the finite additivity axiom
• Strengthens the finite additivity axiom
A3,. . . is an infinite sequence of events,
∪ A2 ∪ A3 ∪If· ·A· )1,=AP
2,(A
A3),.+
1
. . Pis(Aan) infinite sequence of events,
2 + P(A3 ) + · · ·
Countable
then P(A1 ∪ AAdditivity Axiom:
2 ∪ A3 ∪ · · · ) = P(A1) + P(A2) + P(A3) + · · ·
If A1, A2, A3,. . . is an infinite sequence of disjoint events,
then P(A1 ∪ A2 ∪ A3 ∪ · · · ) = P(A1) + P(A2) + P(A3) + · · ·
 Countable additivity axiom

Strengthens subtleties
•Mathematical the finite additivity axiom

Countable Additivity Axiom:

If A1, A2, A3,. . . is an infinite sequence of disjoint events,
then P(A1 ∪ A2 ∪ A3 ∪ · · · ) = P(A1) + P(A2) + P(A3) + · · ·

Mathematical
Mathematical
Mathematical subtleties
subtleties
Mathematical subtleties • Additivity holds only for “countable”
•• Additivity
Additivity holds
holdsonly
onlyfor
for“countable”
“countable” sequences
sequences of events
of events
• Additivity holds only for “countable” sequences of events • The unit square (simlarly, the real line
•• The
The unit
unit square
square(simlarly, thethe
(similarly, real line,
real etc.)
line, is not
etc.) (itscountable
elements cannot be arranged in a
countable
is not
• The unit square (simlarly, the real line, etc.) are not countable
(its
(its elements
elementscannot
cannotbebearranged
arranged in in
a sequence)
a sequence)
• “Area” is a legitimate probability law
•
• “Area”
“Area” isis aa legitimate
legitimate probability
probability law
law onon the
the unit
unit square,
as long as we do not try to assign pro
square,
• “Area”
as is a legitimate probability law on the unit square,
as long
long as
aswe wedo donot
nottry toto
try assign probabilities/areas
assign probabilities/areas to to
“very strange”
“very strange” sets
as
setslong as we do not try to assign probabilities/areas
to “very strange” sets
to “very strange” sets
 Interpretations of probability theor
Interpretations ofInterpretations
probability theory of
• probability
A narrow view:theory
a branch of math
Interpretationsofofprobability
Interpretations probability theory
theory
ow view: •a branch of math
A narrow view: a branch of math – Axioms ⇒ theorems
• –A
ms ⇒ theorems narrow
AAxioms
narrow⇒ view:
view: aabranch
theorems branchofofmath math “Thm:” “Frequency” of event A “is” P(A)
“Frequency”– Axioms
– Axioms
“Thm:”of event ⇒Atheorems
⇒ theorems
“is” P(A)
“Frequency” of event A “is” P(A)
“Thm:” “Frequency” of event A “is” • AreP(A)probabilities frequencies?
“Thm:” “Frequency” of event A “is” –
P (A) toss yields heads) = 1/2
(coin
obabilities•frequencies?
Are probabilities frequencies? P
• Are probabilities frequencies?
n toss yields
•– heads)
– Are(coin
P(coin
P
= 1/2yields
probabilities
toss
toss yieldsfrequencies?
heads) =
heads) 1/2 – P(sole shooter of JFK) = 0.92
= 1/2
e shooter of – JFK) =shooter
(coin 0.92
toss yields – P(a piece of equipment aboard the space shuttle fa
– P(the
P (sole president of .heads)
of JFK)
. . will= be= 1/2
0.92reelected) = 0.7
iece of equipment
– P(a(the aboard
president
piece the of
space
of equipment shuttle
. . . will fails) = 10−8
be reelected)
aboard the space = 0.7
shuttle fails) =a10 −8
• Probabilities are often intepreted as: • Probability models are framework for
ility models are a framework
Betting preferences for describing uncertainty
•– Probability
Probabilities are often
models are aintepreted
frameworkas: for
ing uncertaintyDescription of – Use for consistent reasoning
–
–describing
Description ofbeliefs
uncertainty
beliefs
or consistent reasoning – Use for predictions, decisions
•––AUse for consistent
framework
Betting for dealing
preferences reasoning
with experiments that have uncertain outcomes
or predictions, decisions
Use for
– Rules forpredictions,
consistent decisions
reasoning
•– AUsed
framework for analyzing
for predictions phenomena with uncertain outcomes
and decisions
– PRules
(sole for consistent
shooter of JFK) reasoning
= 0.92
Used
– P
– for predictions
(a piece of equipmentand
aboard the space shuttle fails) = 10−8
decisions
– P(sole shooter of JFK) = 0.92
• Probability models are a framework for
of equipment aboard the space shuttle fails) = 10−8
P(a pieceuncertainty
–describing
 • Are probabilities frequencies? The role of probability th
– P(coin toss yields heads) = 1/2
– P(the president of . . . will be reelected) RealThe
= 0.7worldrole of probability theory
The role of proba
The role of probability theory
• Probabilities are often intepreted
Real worldas: Data
– Betting preferences Real world The role of probabili
The Real world
role of probability theory
– Description of beliefs Data
Inference/Statistics
RealData
world
• A frameworkDatafor analyzing phenomena with uncertain outcomes
Inference/Statistics
– Rules for consistent reasoning Analysis
Inference/Statistics
– Used for predictions and decisions
Inference/Statistics Data
– P(sole shooter of JFK) (Analysis)
= 0.92 Predictions
The role of probability Analysis
theory −8
– P(a piece of equipment
(Analysis) aboard the space Inference/Statistics
shuttle fails) = 10
Predictions Probability theory
• Probability The role
models are a framework
Real world The Predictions
for of probability
role oftheory
probability theory
Predictions Decisions (Analysis)
describing uncertainty
Probability theoryModel building
– Use for
Data consistent
Real world reasoning
Real world Probability theory
Probability theory Predictions
– Use for predictions, decisions
Models building
Data
Inference Data Models building
Models building Probability theory
ory
Inference/Statistics
Analysis Inference/Statistics
Models building

Analysis
Predictions Analysis

Predictions
Probability theoryPredictions

Probability
Model buildingtheory
Probability theory