what is a random variable?
TRANSCRIPT
1
1
Random Variables
2
What is a random Variable?
• RV=Experiment+Measure of interest– Roullette:– Experiment
• Roll the ball (rien ne va plus)– Measure
• A=A red has appeared• B = An Odd number has appeared• C = House winns
3
What is a random Variable?
• Experiment– Flip coins: Heads->+1,Tail->-1
• Measure– A =Fraction of time one is winning– B =Distance between zero crossings– C =Maximal distance to the zero point
4
What is a random Variable?
• Experiment :Diffusion Process
•MeasureA =Maximal distance as a function of timeB=Area/Perimeter
Propagation of messages in a net
5
What is a random Variable?
• Definition:• In a Probability Space (Ω,P), a Random
Variable (rv), is a function:
ℜ→Ω:X
6
Example
• Two players (A,B) roll two dice• Possible random variables:
=Ω Ω∈=
ℜ→Ω
j)(i,for /),(
:
jijiX
X
A
A
We would like to define a probability function
)),((:
jiXPP
A
ℜ→Ω
0 1 P(.)P(.)
↓
),( jiX A
2
7
Example
• Two players (A,B) roll two dice• Possible random variables:
=Ω
Ω∈
<=>
=
ℜ→Ω
j)(i,for ji if 1-ji if 0ji if 1
),(
:
jiX
X
Win
Win
Ω∈−=
ℜ→Ω
j)(i,for ),(
:
jijiX
X
Diff
Diff
0 1 P(.)P(.)
↓
),(or ),( jiXjiX DiffWin
8
Objects that are needed
=Ω
ℜ
+Ω
ℜ→ΩΩ
A ON FUNCTION )),(P( RV a of Prob.
SETA ON FUNCTIONSET ),( Spacey Probabilit
SETA ON FUNCTION : Variable RandomSET Space Sample
jiX
P
X
A
A
Name symbol kind of object
0 1 P(.)P(.)
↓
ℜ∈
),( jiX A
9
Relation between objects
=Ω
),( PΩ
0 1 P(.)P(.)
: ℜ→ΩAX
)),(P( jiX A
0 ℜ
),( jiXA
10
Probability of a Random Variable
• Definition: Set of probabilities associated to the values that a R.V. can take.
),( PΩ
0 1 P(.)P(.)
0 ℜ
) of ( ΩSubsetRandVar( )( ) ( ) ( )
x x
11
Example:Probability of a Random Variable
• A die is thrown twice
1/362/363/364/365/366/365/364/363/362/361/36
12111098765432
0
0,020,040,06
0,080,1
0,120,14
0,160,18
2 3 4 5 6 7 8 9 10 1 1 12
Random Value
Pro
b. o
f th
e ra
nd
om
val
ue
=Ω Ω∈+=
ℜ→Ωj)(i,for ),(
:jijiX
X
12
Example
• Game of Daniel and Nicolas Bernouilli– RV: X(HH…FFH): Fraction of time Daniel is in lead.
ℜ→Ω:X
HHHH HHHH
FHHH H H H H
HFFH FFHH
FFFF FFFF →
L
L
M
L
M
L
=Ω 0 1 X(.)X(.)
X
3
13
Example• Probability of X(HH…FFH):
– We compute P( X(HH…FFH)) for each
)1(1
)(xx
XP−
=π
ℜ→Ω:X
HHHH HHHH
FHHH H H H H
HFFH FFHH
FFFF FFFF
→
→
L
L
M
L
M
L
=Ω 0 1 X(.)X(.)
X
0 1 P(X)P(X)
X
For the derivation of P(X) see Feller or Vélez
Ω∈A
14
Example• Probability of X(HH…FFH):
– RV: Fraction of time Daniel is in lead.
00,02
0,04
0,060,08
0,1
0,12
0,140,16
0,05 0,15 0,25
0,35 0,45 0,55 0,65 0,75 0,85 0,95
Fraction of time a player is in lead
Pro
b. o
f b
ein
g in
lead
a g
iven
fr
acti
on
of
tim
e
)1(1)(
:
xxXP
X
−=
ℜ→Ω
π
15
Example
• Probability of X(HH…FFH):– RV: Fraction of time Daniel is in lead.
• Note that the distribution says that Daniel with high probability either wins of losses
00,02
0,04
0,060,08
0,1
0,12
0,140,16
0,05
0,15 0,25 0,35
0,45 0,55 0,65
0,75 0,85 0,95
Fraction of time a player is in lead
Pro
b. o
f b
ein
g in
lead
a g
iven
fr
acti
on
of
tim
e
)1(1
)(
:
xxXP
X
−=
ℜ→Ω
π
16
Example
• Probability of X(HH…FFH):– Underlying mechanism.– In 20 throws the probability of a run of 5
consecutive heads or tails is much higher than intuition says
00,02
0,04
0,060,08
0,1
0,12
0,140,16
0,05
0,15 0,25 0,35
0,45 0,55 0,65
0,75 0,85 0,95
Fraction of time a player is in lead
Pro
b. o
f b
ein
g in
lead
a g
iven
fr
acti
on
of
tim
e
•Hot hand in basquetball•Highly succesfull wall street trader•Random numbers test.
Methods for RoulleteShannon’s coin game
17
Example
• New R.V. Y: Daniel is in lead more than a given fraction of time.– Meaning: Y=X>f
)arcsin(21)1(
1)()(
:
1
fdxxx
fXPYP
YXY
f ππ−≈
−=>=
→→Ωℜ→Ω
∫
)1(1
)(xx
XP−
=π
ℜ→Ω:X
H H H H H H H H
F H H H HHHH
H F F H F F H H
F F F F F F F F
→
→
L
L
M
L
M
L
=Ω 0 1 X(.)X(.)
X0 1 P(X)P(X)
X
dxxx
fXPf∫ −
=>1
)1(1)(
π
18
0
0,02
0,04
0,06
0,08
0,1
0,12
0,14
0,16
0,05
0,15
0,25
0,35
0,45
0,55
0,65
0,75 0,85
0,95
Fraction of time a player is in lead
Pro
b. o
f be
ing
in le
ad a
giv
en
frac
tion
of t
ime
Example
• New R.V. Y: Daniel is in lead more than a given fraction of time.– Meaning: Y=X>f
)1(1
)(xx
XP−
=π
ℜ→Ω:X
H H H H H H H H
F H H H HHHH
H F F H F F H H
F F F F F F F F
→
→
L
L
M
L
M
L
=Ω 0 1 X(.)X(.)
X0 1 P(X)P(X)
X
dxxx
fXPf∫ −
=>1
)1(1)(
π
f(x)=1-2/pi*arcsin(sqrt(x))
00,10,2
0,30,40,50,60,7
0,80,9
0,05
0,15
0,25
0,35
0,45
0,55
0,65
0,75
0,85
0,95
% of the time a player is in lead
P(%
tim
e in
lea
d)
4
19
Example
• New Rand.Var. Y: Daniel is in lead more than a given fraction of time.– Meaning: Y=X>f
)arcsin(21)1(
1)()(
:
1
fdxxx
fXPYP
YXY
f ππ−≈
−=>=
→→Ωℜ→Ω
∫
f(x)=1-2/pi*arcsin(sqrt(x))
00,10,20,30,40,50,60,70,80,9
0,05
0,15
0,25
0,35
0,45
0,55
0,65
0,75
0,85
0,95
% of the time a player is in lead
P(%
tim
e in
lead
)
20
Distribution of a Random Variable
• Definition:– The distribution function of a discretediscrete random
variable X is defined as:
– With
∑≤∀
==≤=xx
ii
xXPxXPxF
)()()(
m
n
xxxX
AAAX
L
L
,,
,,:
21
21
=
↓
=Ωℜ→Ω
21
Example
• Distribution function of the Rand.Var.: Fraction of time Daniel is in lead
0
0,2
0,4
0,6
0,8
1
1,2
0,05
0,15
0,25
0,35
0,45
0,55
0,65
0,75
0,85
0,95
Fraction of time a player is in lead
Acc
um
late
d p
rob
abili
ty
∑∑≤∀≤∀ −
===≤=
ℜ→Ω
xx iixxi
iixx
xXPxXPxF
X
)1(1)()()(
:
π
00,02
0,04
0,060,08
0,1
0,12
0,140,16
0,05
0,15 0,25 0,35
0,45 0,55 0,65
0,75 0,85 0,95
Fraction of time a player is in lead
Pro
b. o
f b
ein
g in
lead
a g
iven
fr
acti
on
of
tim
e
22
Example• Distribution function of the Rand.Var. : sum
of the values when a die is thrown twice
0
0,020,040,06
0,080,1
0,120,14
0,160,18
2 3 4 5 6 7 8 9 1 0 1 1 12
Random Value
Pro
b. o
f th
e ra
nd
om
val
ue
=Ω
Ω∈+=
ℜ→Ω
j)(i,for ),(
:
jijiX
X
Accumulated Probability
0
0,2
0,4
0,6
0,8
1
2 3 4 5 6 7 8 9 10 11 12
Value of the sum
Acc
um
ula
ted
pro
bab
ility
∑≤∀
==≤=xx
ii
xXPxXPxF
)()()(
23
Example
• Distribution function of a Bernouilli trial until the first success, with probability of success p.– Model of a Bernouilli trials:
• Flipping a coin until a head appears.• Trying to get a connection until the service is given.• Booking a place in an airplane.
– Probability of a success in n trials
ppnXP
Nn 1)1()( −−==
→Ω
24
Example
• Distribution function of a Bernouilli trial until the first success, with probability of success p.
nn
xn SeriesGeometric
n
xn
pp
pp
ppnXPxXPxF
)1(1)1(1
)1(1
)1()()()(
1
−−=−−−−=
=−===≤= ∑∑≤∀
−
≤∀
Note: Complementary of the event,=does appears in none of the n trials
5
25
Example
• Distribution function of a Bernouilli trial until the first success, with probability of success p.
• Coin p=1/2
n
xn
pnXPxXPxF )1(1)()()(
−−===≤= ∑≤∀
Accumulated probability
0
0,2
0,4
0,6
0,8
1
1,2
0 1 2 3 4 5 6 7 8 9 10 11 12
Number of Trial
F(x)
=P(X
<x)
26
Example
• Distribution function of a Bernouilli trial until the first success, with probability of success p.
• Roullette p=1/37
n
xn
pnXPxXPxF )1(1)()()(
−−===≤= ∑≤∀
Accumulated probability
0
0,1
0,2
0,3
0,4
0,5
0 2 4 6 8 10 12 14 16 18 20 22Number of Trial
F(x)
=P(X
<x)
27
Example: Propagation of a rumor
• In a village of n inhabitants, one person explains something to a second, who tells is to a third, etc.
• The first selects one in the n-1 remaining. From then on, all the others select between n-2 (i.e. excludes the previous, and the original)
a) Distribution of the number of times that a rumor is transmited until it reaches again the source
b) Distribution of the number of times that a rumor is transmited until someone get’s it twice. (Exercise)
Taken from C álculo de probabilidades I, R. Vélez & V. Hernández28
Example: Propagation of a rumor
• Distribution of the number of times that a rumor is transmited until it reaches again the source.
– A j (j>2) cannot select the previous or de original
– Note that A1 and A2 cannot select A0
L4
different3-n
3
different3-n
2
different2-n
1
rumor theGenerates
0 AAAAA →→→→
29
Example: Propagation of a rumor
• Distribution of the number of times that a rumor is transmited until it reaches again the origin.
• We define the random variable:– R: Number of times that the rumor is transmited
until it returns to A0
• Scheme of the solution:
)()1()( rRPrRPrRP >−−>==
30
Example: Propagation of a rumor
• We will compute first• Remember:
– A j (j>2) cannot select the previous or de original– Note that A1 and A2 cannot select A0
• A1 and A2 do not count. A1 still has to talk to someone
• Event R>r
2
23
)(−
−−
=>r
nn
rRP
)( rRP >
LLdifferent
2-n 1r
different3-n r
different3-n 4
different3-n 3
different3-n 2
different2-n 1
rumor the Generates 0 AAAAAAA →→→→→→→ +
sequenceother Any AA,A,A 0r43 ANDrR ≠=> L
6
31
Example: Propagation of a rumor
• Distribution function of R)(1)()( rRPrRPrF >−=≤=
≥
−−
<=
−
3r if 23-1
3r if 0)(
2r
nnrF
In a department with n=100
P( R<=r )
0
0,050,1
0,15
0,2
0,25
0,3
0,350,4
0,45
1 4 7
10
13
16
19
22
25
28
31
34
37
40
43
46
49
Repetitions of the rumor
Pro
b. o
f re
turn
ing
to
th
e o
rig
in
32
Example: Propagation of a rumor
• Probability of a given r
21
23
23
- 23
)(323
−
−−
=
−−
−−
==−−−
nnn
nn
nn
rRPrrr
)()1()( rRPrRPrRP >−−>==
In a department with n=100
P(R=r)
0
0,002
0,004
0,006
0,008
0,01
0,012
1 4 7
10
13
16
19
22
25
28
31
34
37
40
43
46
49
Repetitions of the rumor
P(R
=r)
33
Example: Propagation of a rumor
• With n=20
• Note that there is the possibility of loops
P( R<=r )
00,10,20,3
0,40,50,6
0,70,80,9
1
1 4 7
10
13
16
19
22
25
28
31
34
37
40
43
46
49
Repetitions of the rumor
Pro
b. o
f re
turn
ing
to t
he o
rigi
n
P(R=r)
0
0,01
0,02
0,03
0,04
0,05
0,06
0,07
1 4 7 10 13 16 19 22 25 28 31 34 37 40 43 46 49
Repetitions of the rumor
P(R
=r)