Joint Distribution
randomly draw 3 balls
X = no. of red balls Y = no. of white balls
{0,1,2,3} {0,1,2,3}
Prob.
3 2 1 0 X
Prob.
3 2 1 0 Y
(
0)
0.3818 Pr3 12
3 9 =
= =
C C X
0.3818
(
1)
0.4909Pr
3 12
2 9 1 3 × =
= =
C C C X
0.4909
(
2)
0.1227Pr
3 12
1 9 2 3 × = = =
C C C X
0.1227
(
3)
0.004545Pr
3 12
3 3 = = =
C C X
0.004545
0.2545 0.5091 0.2182 0.01818
Joint Distribution
randomly draw 3 balls
X = no. of red balls Y = no. of white balls
{0,1,2,3} {0,1,2,3}
(
0, 0)
0.04545 Pr3 12
3 5 = = = =
C C Y X
3 2 1 0
3 2 1 0 X Y
0.04545
(
1, 2)
0.08182 Pr3 12
2 4 1
3 × =
= = =
C C C Y X
0.08182
(
2, 0)
0.06818 Pr3 12
1 5 2
3 × =
= = =
C C C Y X
0.06818
0.1818 0.1364 0.01818
0.1364 0.2727 0
0.05455 0 0
0 0 0 0.004545
Joint Probability Mass Function
( )
x y(
X xY y)
p , =Pr = , =
( ), ,0 3
3 12
3 5 4
3 × × ≤ + ≤
= −−
y x C
C C C y x
p x y xy
Joint pmf
Example :
0 0 0 0.04545 3
0 0 0.05455 0.06818 2
0 0.08182 0.2727 0.1364 1
0.01818 0.1364 0.1818 0.04505 0
3 2 1 0 X Y
0.04545 0.1227 0.4909 0.3818 Total
1 0.01818 0.2182 0.5091 0.2545 Total
( 0)
PrY= Pr(Y=1) Pr(Y=2) Pr(Y=3)
( 0) PrX=
( 1) PrX=
( 2) PrX=
( 3) PrX=
Marginal Probability Mass Function
( )
=(
=)
=∑
( )
y
X x X x pxy
p Pr ,
( )
=(
=)
=∑
( )
x
Y y Y y pxy
p Pr ,
Example :
( )
, 0,1,2,33 12
3 9
3 × =
= − x
C C C x
p x x
X
( )
, 0,1,2,3 312 3 9
4 × =
= − y
C C C y
p y y
Y
I ndependence
( )
x y p( ) ( )
xp yp , = X Y
X and Y are independent if
for all x,y
0 0 0 0.04545 3
0 0 0.05455 0.06818 2
0 0.08182 0.2727 0.1364 1
0.01818 0.1364 0.1818 0.04505 0
3 2 1 0 X Y
0.04545 0.1227 0.4909 0.3818 Total
1 0.01818 0.2182 0.5091 0.2545 Total Example :
X and Y are related (not independent).0.04505≠0.3818×0.2545 0 0 0 1 3
0 0 0.4444 0.5556 2
0 0.1667 0.5556 0.2778 1
0.0476 0.3571 0.4762 0.1190 0
3 2 1 0 X Y
1 1 1 1 Total
1 0.01818 0.2182 0.5091 0.2545 Total Example :
Conditional pmf
(
)
( )
( )
x p
y x p x y p
X ,
| =
I ndependence
Example :
0.15 0.24 0.24 0.12 40
0.05 0.08 0.08 0.04 20
80 40 20 10 X Y
1 0.20 0.32 0.32 0.16
pY(y)
0.75 0.25
pX(x)
04
.
0
25
.
0
16
.
0
.
32
×
0
.
25
=
0
.
08
0
.
32
×
0
.
25
=
0
.
08
0
.
20
×
0
.
25
=
0
.
05
0
.
16
×
0
.
75
=
0
.
12
0
.
32
×
0
.
75
=
0
.
24
0
.
32
×
0
.
75
=
0
.
24
0
.
20
×
0
.
75
=
0
.
15
0
×
=
Mathematical Expectation
Example : (X, Y) = height and weight of a random man
1 0.18 0.28 0.33 0.21
pY(y)
0.1 0.05 0.03 90
0.24 0.07
0.05 0.02 1.9
0.49 0.15
0.2 0.09 1.8
0.27 0.06
0.08 0.1 1.7
pX(x) 85
80 75 X Y
Joint distribution
( ) ( )(
X =1.7 0.27) ( )(
+1.8 0.49) ( )(
+1.9 0.24)
=1.797 E( ) ( )( ) ( )( ) ( )(
Y = 75 0.21+80 0.33+85 0.28) ( )(
+90 0.18)
=82.15 E( )X =1.797 E
( )Y =82.15 E
?
2=
X
Y
E ( ) ( ) ( ) ( )
5191 . 25
1 . 0 9 . 1 90 07 . 0 9 . 1 85 08 . 0 7 . 1 80 1 . 0 7 . 1 75
2 2
2 2
2 =
+
+ +
+
=
X Y E
( ) ( )
Y EX 2E
≠
(
36X+0.8Y) (
=36×1.7+0.8×75)( )
0.1+(
36×1.9+0.8×90)( )
0.1 E412 . 130
=
( )
X E( )
Y E 0.836 +
=
Mathematical Expectation
(
X Y)
E( )
X E( )
Y E3 +4 =3 +4(
X Y XY)
E(
X)
E( )
Y E( )
XYE7log −2 2+5 =7 log +−2 2 +5
( ) ( ) ( )
XY E XEYE ≠
In general, E
(
X Y) ( ) ( )
≠EX EYIf X and Y are independent, then
( ) ( ) ( )
XY EXEYE =
(
X Y)
E( )
X E( )
YE 2 = 2
(
XY) ( ) ( )
EXE YE = 1
Covariance
Example : (X, Y) = height and weight of a random man
1 0.18 0.28 0.33 0.21
pY(y)
0.1 0.05 0.03 90
0.24 0.07
0.05 0.02 1.9
0.49 0.15
0.2 0.09 1.8
0.27 0.06
0.08 0.1 1.7
pX(x) 85
80 75 X Y
Joint distribution =1.797
x
µ
15 . 82
=
y
µ
x
µ
y
µ (X−µx)(Y−µy)>0
(X−µx)(Y−µy)>0 (X−µx)(Y−µy)<0
(X−µx)(Y−µy)<0
(
)
(
)
[
X
−
xY
−
y]
>
0
E
µ
µ
=
xy
σ
Covariance
Example : (X, Y) = height and weight of a random man
1 0.18 0.28 0.33 0.21
pY(y)
0.1 0.05 0.03 90
0.24 0.07
0.05 0.02 1.9
0.49 0.15
0.2 0.09 1.8
0.27 0.06
0.08 0.1 1.7
pX(x) 85
80 75 X Y
Joint distribution
( )X =1.797 E
( )Y =82.15 E
(
)
[
(
x)
(
y)
]
xy CovXY E X µ Y µ
σ = , = E
(
XY−)
− µ−xµy( ) ( )( )( ) ( )( )(XY =1.7 75 0.1+1.780 0.08)++( )( )( )1.990 0.1=147.745 E
(X,Y)=147.745−(1.797)(82.15)=0.12145
Cov +ve correlated
Covariance
0
>
xy
σ σxy<0
0
=
xy
σ σxy=0
Dependent Independent
Covariance
X and Y independent ⇒ CovE
( ) ( ) ( )
XY(
X,=YE)
=X0EY⇐ ?
Example :
1/ 3 0 1/ 3 0 2
1/ 3 0 1/ 3 30
1 1/ 3
1/ 3 Marginal
0 0
0 1
2/ 3 0
1/ 3 0
Marginal 20
10 No. of dates
Income ($1000)
( ) ( )
3 40 20 2 30 0 10 0 3
1 × + × + × = =
XY E
( ) 32 2 3 1 0 3 2× + × = =
X E
( ) (10 20 30) 20 3
1 + + =
=
Y E
( ) 20 0 3 2 3 40 ,Y= − × =
X Cov
( ) ( ) Pr( 0, 10)
9 2 10 Pr 0
PrX= Y= = ≠ X= Y=
X and Y are not independent!
Correlation Coefficient
Magnitude of σxydepends on the scale of X and Y.
(
aX b cY d)
acCov(
XY)
Cov + , + = ,
Standardized by standard deviations of X and Y :
(
)
( ) ( )
XVarY VarY X
Cov ,
(
)
= =CorrX,Yρ Correlation coefficient
1
1≤ ≤
− ρ
(
aX b cY d)
sign( )
acCorr(
XY)
Corr + , + = ,
X and Y independent ⇐⇒ρ= 0
Correlation Coefficient
Example : (X, Y) = height and weight of a random man
1 0.18 0.28 0.33 0.21
pY(y)
0.1 0.05 0.03 90
0.24 0.07
0.05 0.02 1.9
0.49 0.15
0.2 0.09 1.8
0.27 0.06
0.08 0.1 1.7
pX(x) 85
80 75 X Y
Joint distribution
12145 . 0
=
xy
σ
( )X =1.797 E
( )Y =82.15 E
( ) ( )
X2 =1.72(0.27)+( )
1.82(0.49)+( )
1.92(0.24)=3.2343 E ( )X =3.2343−1.7972=0.005091Var
( ) ( )
Y2 = 752( )0.21+( )
802(0.33)+( )
852(0.28)+( )
902(0.18)=6774.25 E ( )Y =6774.25−82.152=25.6275Var
( ) ( ) ( )XVarY Var
Y X
Cov ,
=
ρ ( )( ) 0.3362
6275 . 25 005091 . 0
12145 .
0 =
Slightly +ve correlated
Linear Combination of Random Variables
(
aX bY)
aE( )
X bE( )
YE + = +
(
aX bY)
aVar( )
X bVar( )
Y abCov(
XY)
Var + = 2 + 2 +2 ,
Example : X = husband’s income Y = wife’s income
( )
X =20E E
( )
Y =16( )
X =60Var Var
( )
Y =70 Cov(
X,Y)
=49Total income : S = X + Y
( ) ( ) ( )
S =EX +EY =20+16=36 E( )
S Var( )
X Var( )
Y Cov(
XY)
Var( )
S ==60+70++2×49=+2282 ,Var σS= 228=15.1
Linear Combination of Random Variables
Example: investment
0.4 10%
0.3 0.2
0.1 Probability
20% 0%
-10% Return on Stock
0.2 10% 0.6 0.2 Probability
8% 6% Return on Bonds
( )
S =9% E( )
B=8% E( ) 2
% 89
=
S Var
( ) 2
% 6 . 1
=
B Var
1 0.3 0.4 0.2 0.1 Margin
0.2 0 0 0.1 0.1 10%
0.6 0.2 0.3 0.1 0 8%
0.2 0.1 0.1 0 0 6%
Margin 20% 10% 0% -10% B S
( ) (
SB =−10)( )( ) ( )( )( )
6 0+0 6 0++( )( )( )
20 10 0=64 E( )
S,B =64−9×8=−8Cov
( )
2% 8 ,B=− S Cov
(
p)
B pSR= +1− Portfolio
( )
R pE( ) (
S p) ( )
EB E( )
R= p(
+1p−)
E( )
R=9 +p81− E =8+( )R pVar( ) (S p)Var( )B p( p) ( )CovSB
Var( )= 2 +(1−)2 ( +2)1− ,
p p p p R
Var
( )
=89 2+1.61− 2−16 1− 6 . 1 2 . 19 6 .106 2− +
= p p
R Var
( )
8.09 0.857609 .
0 = R=
= ER σ
p
Population and Sample
PopulationSample
2
,σ
µ Population parameters
2 ,S
X Sample Statistics 2 ,S
X 2
,σ
µ
Inference
Sample
Population and Sample
Population6
= µ
8
2=
σ
Sample (with replacement)
Sample Mean Sample Variance
{
X1,X2}
22 1 X
X
X= +
{
(
) (
2)
2}
21 2
1 2
1
X X X X
S − + −
−
=( )
2
2 2 1
2 X X
S = −
……….
2 0 Prob = 1/25
3 2 Prob = 2/25
Population and Sample
0.08
9 10
8 7 6 5 4 3 2
0.04 0.12
0.16 0.2 0.16 0.12 0.08 0.04 Prob
X
Sampling distributions
0.08 0.16 0.24 0.32 0.2 Prob
32 18 8 2 0 2 S
( )
X =2×0.04+3×0.08++10×0.04 E( )
X =6=µE
( )
2 =0×0.2+2×0.32+ +32×0.08 SE
( )
2 =8=σ2 SE Unbiased
( )
X=4⇒ Var( )
X =2Var Var
( )
S2 =86.4⇒ Var( )
S2 =9.295Very Simple Random Sample (VSRS)
{
X1,X2,...,Xn}
(very simple) random sampleEach X drawn from same population (distribution)
X’s are independent
( )
X =µE
( )
n X Var
2
σ
= SE= Var
( )
X =σn standard errorExample : Final examination scores µ=71.8 σ2=195.2
For a VSRS of 4 students : E( )X =µ=71.8 7.0 4
2 . 195 =
= =
n SE σ
For a VSRS of 16 students : E( )X =µ=71.8 3.5 16
2 . 195 =
= =
n SE σ
Sampling Distribution
Population{
X1,X2}
0.08
9 10
8 7 6 5 4 3 2
0.04 0.12
0.16 0.2 0.16 0.12 0.08 0.04 Prob
X
Sampling distributions
0.08 0.16 0.24 0.32 0.2 Prob
32 18 8 2 0 2 S
VSRS
Sampling Distribution
Population
{X1,X2}
VSRS
Dist. of 2
S
X S2
Dist. of X
Normal Population
Population ( 2)
,σ µ
N
µ
{X1,X2,...,Xn}
VSRS
Dist. of ( 2) 2
1
σ
S n−
2 1
−
n
χ
Dist. of X
n N
2
,σ
µ
µ
X 2
S
Normal Population
Example : measurement X , true value µ
X ~ N(µ, 0.01)
(
)
> − = > −
1 . 0 05 . 0 1 . 0 Pr 05 . 0
Pr X µ X µ =2(1−Φ( )0.5)=0.617
Take 10 measurements independently. (VSRS with n = 10)
10 01 . 0 , ~N µ X
(
)
> − = > −
10 1 . 0
05 . 0 10 1 . 0 Pr 05 . 0
Simple Random Sample (SRS)
{
X1,X2,...,Xn}
simple random sampleSampling without replacement from population (size N)
Equal probability for each possible sample
( )
X =µE
( )
n N
n N X Var
2
1 σ
− −
= ( )
n N
n N X Var
SE σ
1
− − =
= standard
error
1 1≤
− −
N n N
Finite population correction factor
Simple Random Sample
Population6
= µ
8
2=
σ
Sample (withoutreplacement)
Sample Mean Sample Variance
{
X1,X2}
22 1 X
X
X= + ( )
2
2 2 1
2 X X
S = −
……….
6 8 Prob = 1/10
3 2 Prob = 1/10
4 8 Prob = 1/10
Simple Random Sample
9 8 7 6 5 4 3
0.1 0.1 0.2 0.2 0.2 0.1 0.1 Prob
X
Sampling distributions
0.1 0.2 0.3 0.4 Prob
32 18 8 2 2 S
( )
X =3×0.1+4×0.1++9×0.1 E( )
X =6=µE
( )
2 =2×0.4+8×0.3+ +32×0.1
S E
( )
S2=10≠σ2 EUnbiased
( )
X=3⇒ Var( )
X =1.732Var Var
( )
S2 =88⇒ Var( )
S2 =9.38Biased
Sampling Distribution (SRS)
Population
{X1,X2}
SRS
Dist. of 2
S
X S2
Dist. of X
small n
X
moderate n Very large n
Central Limit Theorem
Arbitrarypopulation
(mean µµµµ, variance σσσσ2)
{
X1,X2,...,Xn}
VSRS
Central Limit Theorem
Arbitrarypopulation
(mean µµµµ, variance σσσσ2)
{
X1,X2,...,Xn}
VSRS
( )
→∞→ −
n N
n
X L
as 1 , 0
σ µ
( )
− Φ ≈ ≤
n c c X
σ µ Pr
For large n
( )
− Φ = ≤c cσµ
X ?
Pr
Normal Approximation
Example : Monthly income
0.05 20000
0.03 30000
0.01 40000
0.01 0.1
0.15 0.25 0.4 Prob
60000 15000
10000 7000 4000 Income
( ) (
=4000)( ) (
0.4+7000)(
0.25)
+ +(
60000)(
0.01)
=EX
µ=E
( )
X =9250µ
( ) (
2 =40002)
( )
0.4 +(
70002)
(
0.25)
+ +(
600002)
(
0.01)
X
E
( )
2 =155150000 XE2
( )
29250 155150000−
= =VarX
σ2=
( )
=69587500 XVar
σ
( )
=8342= VarX
σ
For a VSRS with size 100
100 8342 , 9250 ~
2 .
N X
(
)
1 (0.90) 0.184110 8342
9250 10000 1 10000
Pr = −Φ =
− Φ − ≈ >
X
( ) 1 (0.09) 0.4641 8342
9250 10000 1 10000
Pr = −Φ =
−
Φ − ≈ >
X
(
10000)
0.1 0.05 0.03 0.01 0.01 0.2PrX> = + + + + =
Normal Approximation
Normal approximate binomial
( ),π
~bn
Y nlarge
(1 ) ( )~ 0,1
. N n
n Y
π π
π
− −
Example : Y~b(35,0.25)
( )
Y =35×0.25=8.75E Var
( )
Y =35×0.25×0.75=6.5625(
)
− Φ −
− Φ ≈ ≤ ≤
5625 . 6
75 . 8 7 5625 . 6
75 . 8 15 15 7
Pr
(
7 Y 15) (
2.440) (
0.683)
Pr(
7≤Y≤15)
≈Φ0.9927(
−1Φ0−.7527)
0.7454Pr ≤Y≤ ≈ − − =
By exact calculation,
(
7 15)
(
0.25) (
0.75)
0.8018 Pr15
7
35
35 =
= ≤
≤
∑
=
−
y
y y
y
C Y
Normal Approximation
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
0 5 10 15 20 25 30 35
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16
0 5
7
10 15 20 25 30 35(
)
(
7 35,0.25 15)
PrPr(7≤≤Nb(8.75,6.5625≤)≤15)
6.5 Continuity correction 15.5
Continuity Correction
Approximate discrete distribution by continuous distribution
(
X≤c)
Pr Pr
(
X≤c+0.5)
(
X<c)
Pr Pr
(
X≤c−0.5)
(
X≥c)
Pr Pr
(
X≥c−0.5)
(
X>c)
Pr Pr
(
X≥c+0.5)
Example : Y~b(35,0.25) E( )X =8.75 Var( )X =6.5625
(
)
−
Φ −
−
Φ ≈ ≤ ≤
5625 . 6
75 . 8 5 . 6 5625 . 6
75 . 8 5 . 15 15 7
PrPrPr
(
(
77≤≤YYY≤≤1515) (
)
≈≈Φ0.99582.635−) (
(
−1−Φ0−.81000.878)
=)
0.8058By exact calculation, Pr(7≤Y≤15)=0.8018
Normal Approximation
Normal approximate Poisson
( )θ ℘
~
Y θlarge Y ~.N( )0,1
θ θ −
Example : Y~℘( )30
(
)
−
Φ −
−
Φ ≈ ≤ ≤
30 30 5 . 23 30
30 5 . 39 39 24
Pr
(
24 Y 39) (
1.734) (
1.187)
PrPr(
24≤≤YY≤≤39)
≈≈Φ0.9586−−(
1−Φ0−.8824)
=0.841By exact calculation,
(
)
0.8391! 30 39
24 Pr
39
24 30
= =
≤
≤
∑
= −
y y
y e Y
By normal approximation with continuity correction,
Normal Approximation
Normal approximate Chi-Square
2
~ r
Y χ rlarge ~ ( )0,1 2
. N r
r Y−
Example : 2 80
~χ Y
(
)
−
Φ −
−
Φ ≈ ≤ ≤
160 80 39 . 60 160
80 58 . 96 58 . 96 39 . 60
Pr
(
60.39 Y 96.58) (
1.311) (
1.550)
Pr(
60.39≤Y≤96.58)
≈Φ0.9051(
−1Φ0−.9394)
0.8445Pr ≤Y≤ ≈ − − =
From Chi-Square distribution table,
(
60.39 96.58)
0.9 0.05 0.85Pr ≤Y≤ = − =