MA 20104

Probability & Statistics

Swan and Khare

Slot A 3 .

( AG t MI t IM )
 :

Randy
An experiment whose outcome is

Experiment
Unknown .

Ex .

Tossing a coin

time in
Waiting a
queue

samphs-paa.ch
)

Given a random
experiment , a set I
all possible outcomes is a
sample
Space
.

D= EH , T ) -

tossing coin

die
R -
- El , 2,3--14,5 ,
-

Rolling a

time in
waiting
a
Co
r
)
-

= a
,

green

trent : Subset 8 the sample space

.
 discrete
/ EH ,T3 E 1,2 63,1N
Sample
.

,
-

,
,

spaces - continuous

[ op )
,
Ca , b)

events
Assignment f probability to .

cage : discrete I finite I all outcomes are

equally likely .

A be event Relative
Let an .

f.
#I g regnency
HAI
assignment
=
.

# r

case .
discrete / finite / all outcomes are

Not equally likely .

Run the experiment n number 8 time .

By fa denote the relative fngnecy

times
no 8 event A occurred in n
runs

fa
-

= -
n

BCA ) FA
=

nhjm ,
 needs
for
L
only kcyents
-

Obsa : certain
-
an

to
assign probabilities .

for die
instance in of rolling
case a we
,

need tell Hal Pls )

to assign only , , ,

all other
Pk )
,
PK ) . Then
probability 8
events are automatically assigned .

How does this automatic assignment

=

? ?
happens
certain laws
- Because are
intuitively assume

of probability .

-
We will formalize these laws / these

intentions into the axioms and

propose
an axiomatic definition 8 probability .

Why we need axiomatic definition :

it works in all possible scenes .

it matches with the ideas stated previously .

about probability assignment .

 subset 8
Event : A r .

Union / intersection of events is also

an event .

Definition : T - field .

A collection of sets F is called

5-
algebra if

i ) AEF then ACE F

,

Ii ) Ai Az E F then
'

Az
- .

,
-

, ,
.

iIAi E F and Ai EF

EI .
R -

- EH ,
T }

F = Er = E to EH }
, ,
ET } E Hit } }
Ex : re 91,2 , 3,4 , 5,6 }

F = Er
 Definition : Lets be the sample space
for a random experiment .
let F be

associated
the f-
algebra with SL .

is
The
probability P :
a measure a

function on F to IR C p : I → IR )
such that

i ) PIA ) > o t AEF

ii ) Pfe ) = I

for
iii
) If Ai are mutually disjoint
i -4,3 - - -

( Ain Aj =
4 for itj )
PL Ai ) -

I ,
PHI )

Properties :

PIA I
'
i
) =
I -
PIA )

A UAC = r

'
Pcr )
'

NAVA ) = PIAITPIA ) =

in
because A4A° are distant
g) plot ) -_ o

A r result
take
-
-
in previous .

A CB PLATE PCB )
3) If ,
then

µf B=HulB
disjoint
union

PIB )
,¥tPlB
Pp
; so

4) PCAVB ) = PLAIT PIB ) - PIAAB )

FiT AaB
-

'
AUB
=fAnB ) ULANB ) UCACNB )
disjoint .ae
←
PCAUB ) = Plan Bc ) + Pl Acn B) t PLAN
B) ti )

Now notice ,

'
A = ( AaB ) U ( AaB )
←
disjoint
union
'
PIAS = Plans ) + plan By

Plan B
'
)= MAI -
Pl AaB ) -127

similarly , PLAT B) =p LB ) -
PLA AB ) - Cs
)
From C
17,121 ,
(3)

RAUB ) = PLAIT PIB ) - Plan B ) a

PCAVBUC ) PLAIT Pl B) PCC ) PIAA B )

5) t -
=

PC Bnc ) -
Plan t PIAA Bnc )
similarly using induction
,
prove
PIA , U Az . . .
VAN ) -
-
Eixample
Experiment fair coins
-

i
)
: Toss 2

D= { IHH ) , CHT )
,
CTH ) C TT ) }
F = 2B
PIHH ) k
Assume : PC HH ) =
I =

HT )=
PC
at PC HT ) = Pz

PITH )
I PITH ) Pz
=
=

PITT )
tf PITT ) A Hitz
=
=
I - -

OSP , ,k ,
BCI
of Past
-

theorem I : Let A, c Azc As C -

- - -

C 8 sets )
sequence increasing .

and then
A =
VAI
IT

PCAI = him Plan )

now
 As

PI B ,
=
At

Az = Az AAE

Bz =
Az A AE
:

An Ants
Bn =
n

II. Bi =

II.
Ai

A =

II. Ai = Bi
 .de?BiI=EIPCBi
RAD
=P )
°

Let {Bi ,
Bz
,
- - .

, Bn } be a
disjoint
cover of r .

)
d )
[
( Bi ti ,j
Bj 1,2 n
n = =
,
.
-

and
ii. Bi = r

Then for any event A CAEF )

PCA ) =P ( AaB , )tPlAaBd t . . .

t PCAOB n )
 Conditioning .

t PCB ) >
f
o
Let A ,
BE s . .

Conditional probability g A
given
B .

PCA 1B ) = NAIB )
ng
PCB )

fair die
R Rolling
: a

top face of
A 2 the the die
appears
: on

B number the top

face
: even
appears on

To the die

main =3
It
?
:& ,

=*÷÷¥±i÷÷÷÷
It I AaB )

probability .
 B) PCAIB ) PCB )
rule
Plan = - - -

product
41 ) of probability

If PLA ) > 0

PC BIA ) =
NAIB )
PLA )

PLAN B) = PIBIA ) .
PIA )
427

From a ) & (2)

PCAIB ) PCB ) = PCBIA ) PIA )

::::::÷÷÷÷
Consider { Be
,
Bz
,
-
-

,
Bn } disjoint cover fr .

&

! )
of for
Bin Bj =

ifj Assume that

! Bi =
r
no Bi =
0

For any At F

A = ( AaB , ) U C An Bd U -
-

Ulan Bn )

Idisj.int
union

plate pl AaB , ) t Pl An Balt - - -

t Plan Ba )

,) PIAIBIPIB.lt
-
- PIAIB PIB It ,
-
- .

.
.
-

t Plott Bn ) PCBs )

IPIAI-iIPIAIBIIPIBTY-totafpkwmae.in ,
-

pcpsgla )p1Bi)
PlAnlBglPlB# ) Baye 's

EPIAIBI
= ←

rule .

E- i
independence
Conditional A given B ( Pbl 'D
probability 8

PLAID ) = NAIB ) '

PCB )

3,43 all the outcomes

EX :
1=91,2 ,

are
equality likely .

A =
{ 1,2 } ,
B =
El 33 ,
,
C -
-

91,4 }
A ,
= El , 2,33 ,
Are -
- fi }

AB
,
=
{ 1,433 ,
Bz = El , 2,43

IA )= PIAL ) =L
PIA , ,
Is 4

I Biltz ) PIB )=
}
P =
,
Ig

PLA ) =
I PIAIB ) -

Iz
PCB )= PCBIA )
I =
Iz

Plan B) PCA) PIBIA )

definition 8
= .

IPIAABJ-piag.pl# -
←
independence .
 events
Pairwise independence 8
-

Plan B) = PIA ) .
PIB )

If is dear that AB are pairwise

and and Bic
independent so are Ac
in the previous example .

rCdAnr
-

the
dkef.ohm.FI
-

mutual independence

PIAA Bac ) = PLA ) PIB ) p

be two events from

Definition : Let A ,
B

Cd ,
FA ) .
we called A- dB pairwise
independent if
Plan B) = PIA ) - PCB)

Definition ;
Let A ,
B
,
C be three events from

( r
,
F ,
P ) .
We call A ,
B, c mutually
independent if

PIAAB7-PIAI.plrdjpbnd-PIB7.PH i

PIAAC ) = PIA ) .
PCD ; PIAABnct-PIAI.pl Bl Pk )
 mutual
we can
generalize the definition q
independence to the events As Az ,
.
-
.

,
An
,

in an inductive way .

sub collection I Ay An containing

Any
- .
.

-
,

at least 2 events and almost nt events

is mutually independent
-
PIA ,
AAH -
-

nan ) =P HAP LAD . . -

plan )

E Setting up probability space in case 8

repleted random experiment C with only 2

possible

outcomes
) .

Eg :
tossing
a win ) R
r -
-
fo B ,

F -

-
I = { 0, r
, Sol ,
3
Pho 3) = I -

b ;
P ( fi } ) =p

Set when
up probability space one

with the
f times
' '

repeats R n number

assumption
that
repetitions are independent
if each other .
 For this
compound experiment
D= {9, Yz Tn I Yi uh in -
-

,
- -
C- 90,13 }
"
# D= z

F = 2h power set 8 r

Probability assignment .

th
Probability of be successes in

repetitions at specified -
locations .

Conoid er an element wer

W =
y , Yz . .
.
Dn

where 91,92 9k =L Theta Yn O

;
- - =
,
.
.

, ,

( W = I I .
.

-
- I O 0 -

on
-
o )
n h times
is
times
-

Ai = Success at in toss for i -

1,2 ,
- -

is

P ( w ) = p( bi -
- -

bn ) =p ( Il
- -

I 00 - - o
)
a- I

mutual Plain Azn A Ann After rain)

G
=
.
- -
.
.

independence
= play .
. . .
pl Aa ) Plain ) -
- -
PHI )

pk pyn
h
-

=
C , -
NNOW if is interested in computing
probability
one the

8 the event that

exactly
tosses
k number of successes in A coin .

a
bk pj
-

(I) a
= -

i'
Example : A box contains o balls .

6- red 60%
4 blue
Hob
-

)
Casely Pick a ball and replace it in

)
ball
the box and pick a
again .

A ,
= red ball is picked in the first
trial .

Az = blue ball is picked in the

Second trial .

PCA ,
I
Ai ) = PC Are) =
Fo = O -

This is called
process
as
,
with
Sampling replacement .
 case Ca ) Pick a ball ; do not
put it back

& ball
in the box pick a
again
I Sampling without replacement ) .

the first
A, = red ball is picked in

trial .

the second
Az = blue ball is picked in

trial .

PIAZIA ,
)=
Ig
t
play = Fox 's t To
.

}
not
Thus Aid Az are
independent .

f
'
'
Increase n and keep the
proportions
the red balls and blue balls constant ,

€ )
Aid Az start behaving more and more A

he independent events :
 Uniform
probability spaces .

Experiment Picking up : a number randomly

from a given set .

cast ) :

Given set is finite and discrete .

{ 1.3 - - -

,
n
} -

-
r

with
Probability subsets f
r
8 two

This is the
equal site
=
be equal .
C

uniform probability principle )

Here we associate
size to cardinality
of the set .

that
As a
consequence ,
we can see

The Singleton sets are

assigned probability
I '

( these are called discrete uniform

spaces as

probability spaces )
 CH : The
given set r is finite and

an interval f R .

In
particular ,
take r = Co ,D

Principle 8 uniformity : Pick two subsets

-

should
"

f
"

r with
equal size ,
then they

have same
probability .

Take an interval Ca ,
b) E Co , D
la 's ) is associated with the
site g ,

which b
length 8 Ca ,
b ) is -
a .

Probability assignment by obeying principle of

uniformity .

two intervals f same length are

Any
to their
assigned with probability equal

length .

for Ca b) EBD
PCA) b ) b a
any ,
-

=
,
The so -

algebra F ,
here is the set f

subsets Coil ] which generated

f are

as countable unions and intersection J

intervals .
C Borel r -

algebra )
This is the continuous uniform probability
space
singleton set
Remand Note
:D that any
zero
has probability
.

Ears E Coil ]

a) Consider the set f rational numbers

in Co , D denoted as Q n Cosi ) .

Then from CD and the fact that

Qnfod ) is countable
,

Plan Coin ] = o

A
 A machine contains 4
E± components
parallel with 3
in
o I o 2,0 o 4
-
. .
.

, ,

as their probability 8 failures

respectively .
The machine fails if all

the
components fail simultaneously Nole that
.

The failure f machines is

independent
.

what is the that

Then
probability the

machine once started with

not fail ? ?

that the i
Let Ai : event component
fails .

Ash ALAA nay :

machine fails
,

I -
P [(A ,
a Azn Ash Aa ) ) =P( AAAAA ,
Add

= I - PIA , ) . NAINA ,
)plAa )

= I -
fo
i ) ( o 2) to 3) 10.4 )
. .
= I - 0.0024
 hey

E±
A rocket
engine fails if one component

fails .
The probability of failure I this

Soho order to
key component is .
.
In

increase the success

proof ability f the

rocket engine this

an
assembly f key

components in
parallel is proposed so

that the engine

fails if all these key
What
components fail simultaneously .
is

the minimum number f key Components

are
required in the
parallel assembly
So that the engine has 99% of
??
success
probability
"

I -
Co05 )
.

go.gg
+
independence
 A probability
E course on is a
very
and students allowed
famous course are

to register after the teacher consents .

If is observed that 20% times after

obtaining the consent ,

student fails to

register for the course . there an only

too seats available in the class .
If a

teacher consented 102 students what

,
is the that all the students
probability
will be accommodated in the class ? ?

" 2 to '

f- ( o 8).

-
102×10.8 ) o . 2
EE In class too students
.

f
30 -
Mi

40 -
IM

30 -
AG

707 mi students are dual degrees

of
% IM dual degree
↳
if
dual
350% 8 AG degree

EI .
3
printers

DP
DP
DB .

Printer Pc
gets 50 To jobs

30 %
p
2

- -

has its
Printer P
, o.is as
probability
8 failure
,
Pz o .

I as its probability 8
and
printer Pz o .
2 as its probability
8 failure .

If a
randomly selected printing job is

a failure what is the

probability
,

that it was printed by printer B ? ?

F : failed fab
PLFIP , ) = o .

15 PIF 1Pa ) =
0 .

I ,
,

Pfftpz ) =
0.2

PIP , ) = o .
5 Pfk ) =
o -3 Mfs ) -
-
0.2
,
,

compute plots IF )

Plf 1B ) PCB )
PIBIF ) =
.

c- H
EPLEIPI )Plei) T
=
0-332
 2 thrown What is
Ex .
fair dice are .

obtain
the probability that gone
will

two sixes given that

416
Done has I
six already appeared .

e) sum 8 the two faces is greater

than six .

( Ya ) ,

EE A- sixes
appear
= 2

D= at least one six has

appeared
.

r={ Cfl ) ,
Cl
, 2)
,
-
-
-

,
11,61 ,

( 21117,0421 ,
.
-
-

.
Csg
C

}
 Random experiment
Cr ,
F ,
p)

A random variable
Randomabh :

X is a function from r to IR

if Ews ,
XCW ) - n } C- F .

"

it NGIR
, { w : Xlw ) -_ a
} = X In )

inverse
Er
image be .

{w/xlw1=a } e y
Plx -

-
a ) = Pfwlxlw ) -
.
n }
 3h
fair
tossed
Expt :
coins
independently .

D= { HHH
,
HHT
,
HTH ,
TH H
,
TTH ,
THTITTH ,

TTT}

f- =
2h

PCHHH ) =
Izxtzxtz =
I
× : no . of heads in 3 tosses .

µ
r

H
x
HH

HHT

HTT
THT

*
T 3

2=2
,
x' (2) ={wEr/ Xlw ) =
2
}

=
{ HHT
,
HTH
,
THH }

PIX = 2) = p { WE rlxlw )=2 } =

PEHHIHTH ,THH }
=
I
8
 k -
-
8- I

Plx = se ) = PCX = 8. 7) = O

*
Range of x :
{ see art Xin ) Do }
* A random Variable X is called a

discrete random variable if Range f x

finite infinite
is or
countably .

In the previous example

Rx = { o
,
I
,
2,33 and hence X is a

discrete random variable

A real valued function

Definition :

fin ) PIX -
- a ) is called as discrete
=

density or
probability mass function C pmf )

of the discrete v.v .

X .

for all
ffse ) =
I see O foe )=o
8 other

= I R = I values
8 8 a .

2
= I N =

I 2=3
=

8
 ←
graphical

f
x x

sT'- /
representation

/ / 8 pint

← tabular

-
|An/%#-↳µyg/ representation
se

-
8 pmf

Exit Bernoulli

coin
Expt Tossing
a
:

r -
- { H.TT} PCH ) =p ; PCT ) -
-
I -

Xl Heads ) =L XC tails ) = O
,

PIX -
-
o ) = I -
p PIX -
- 1) =p
,

}
f-Cn ) l -

p re = O
;
-

f
Pim
-
.

= p ;
n - I

= O
otherwise

X N Bernoulli ( p) Rx .
- fo ,
'
}
+
parameter
Ex :
Binomial Cpj
Bernoulli trials
Expt : n
independent are

performed -

trials
X : number of successes in n

Rx = fo ,
1,2 ,
-
-
.

,
n
}

flu )= PIX = a )

" " -
"

fly ) (f) p (I P ) 1,2

-
a -
o -
n
,
-
-

= , ,

otherwise
= O

× N Binomial ( n
,
t )

Ex : Geometric

Perform trials Ip) until the

Exp indep .
Bernardi

first Success f

re ,
FS
,
FFS
Ef
,
F # FS
,
.
-
.

]
rat . terrific .

n 's
X :
The number of failures before the

first success .
 Rx = { 0,1 ,
2
,
-
- .

3 discrete r . v .

se

f- he ) p a p)
o
1,2
.

se
-
-

,
-
= - . ,

= o
otherwise .

for random Pkr ) I

that appt
-
* Note any
.
.

Let X be a random variable defined mr .

Let Rx be the 8 this random

range

Variable .

Rx =
EM ,
Nz
,
.
.
.

,
an
,
-
. .

C variable can ) .

disown
"
"
E.
"
for
"
nil =
-

g
}
ni .

= E PEW @
:X I -
-
ni }
ni C- Rx

= I

Random
exp 't .

→ r ,
I ,
P → → * pint
 Properties of pmf
foe IR
i) ffa ) > o
E

Ese fin ) > o

} is finite countable
ii
)
: a or

= Em ,
Nz ,
- - .

,
en
,
. -
3 Set .

iii ) Efcni ) = I
ni

function if which tsatisfies these three

Any
called discrete
properties is as
density or

pmf .

Ex : Uniform pmf
An ) =
I RE f see , Nz
-
-

,
Hs }

= o
otherwise

Ex : Geometric pmf
ffa ) = PCI -

pj se = o
, 1,2 ,
.
-
.

= O otherwise
 Cd 't )
Ex : Negative Binomial
"

Atal ) fp )
'

Pd ( 0,12
-

-
n -

-
. . .

,
fin ) =

otherwise
= o

Cexpt :
performing independent Bernoulilp )

trials unit 2 no .
of successes are

observed I x : no .
of failures before
.

successes
-

a
for
.

2=2

{ HH
,
HTH
,
TH H
,
. - -

Ex :
Poisson density Cd ) Cx
> ) o

I a

fln ) =
C A see 0,1 ,2 ,
-
- .

x
!

= o otherwise
 Special examples 8 r . v. s .

Ex .
.
Constant random variable

A WER
XCW ) = C

c) # c)
0
PIX -_ = I PIX =

I "
#

EI Indicator random variable .

let A EF
WEA
XA I w )
t
-
I

'
=o Aw E A

E± Hypergeometric distribution .

of r
objects
Population .

Type I objects r
,

Type II
objects
r
r
rz
- =
,

let a sample of size n is chosen from

this population C near )
 ( ne re )

X : number of
objects of Type
I in the

random
sample .

¥t#÷⇒
:*:

( LIL )
PK -

-
a) =
-
M -
-
0,1 ,
2
,
- -
-

in

Tn )
= o otherwise

Computations
-
with
pmf :

( 1,7 ,
p )

× : discrete r . v .
with pmf fix )

Event : X It for some real no . t .

= Ew : xcw ) EHS
t

U Ew xlw )=I }
=
:

pilot
?
i= -

PHET ) =
Ew : xiwy.si?)=!IIPlHEi)
,
 Cumulative distribution function .

discrete foe )
random variable x with pmf
For
every
real number TE R
,

define

:I÷I÷
Fct ) PIX Et ) '

}
=

= tens function

Ex : XN uniform ( s -
- lo
)
flu ) I 2 9
= a =
0,1 , ,
.
- -

,
10

= o otherwise

dearly for too ,

fCtI= o

Flo )= flo ) = I
( O

for any costal Fft ) -_ I

⇒
to

FC D= floltffl ) go
=

for 1st ez ffCt7= Z

any ;
to
 it -

-0 iii.
Yo
.

f function
CDF : Cumulative distribution
fin )
let x be discrete r . v . with pmf .

For TE IR we define
any

Effie
Fft ) =
)
zest

function from IR to IR
clearly f is
.

a
,

F function
.
is a non -

decreasing .

for FADE Fitz )

tietz ⇒

• FC -83=0 ,
fld ) = I

¢ 11

lim Flt ) Fct ) I

= o Iim =

t -

s - N
too
 The of F like staircase
.
graph is a

at the
with jumps occurring exactly
points which are in the of
range
X .

. .
Right continuity 8 F

Iim Fln ) =
Fla ) t AER

a) at

EI X N geometric Cp )
N
1,2
fine x o

p ( p)
,
- .
- -

)= I
-
-
,

=
o otherwise

F It ) = 0 t too
Et
a
t
F- It ) = I PCI -

p) > o

N=O

pgft ) ti

]
g-
[
I
=P
-

It ]t I

I (I p) At
Fit ) = > o
-

-
 NE Rx
Events
for this
geometric
r .
v -

×< x and X 7N
pcx > n ) = I -

Plan )
I ×< a u x > near

÷!
:÷
9
I PUSH D disjoint
-

f
= -

Fla
U
= I -
-

I )
' " "

:
' is i

n . > .
,

Love :
p Cx > ntm ) x > n ) =
PCX > m )
-
E less
11 memory property

p ( X > ntm n X > n )

-

P CX 7 n
)
=
plm )

PCX >
NT
-

It T H
,
TT H TT TH TTTTH -
.

,
-

,
, ,
 R random
-

eypt .

r
, I ,
P

discrete r - r .
X

f floes
p.m
.

c- D. F .
FCK )

× is a random variable Car ,

F ,pT
Probability of the event Xen

F Cn ) = PCX Ese ) HIER

This function F is called as cumulative

distribution function Fcn ) of rv .

X .

Cfda ) )

Properties of CDF .

I . of Fxfd ) I I A NER

2. him Else ) =L
ATTN

S .

lim Else ) = O
NS - N

4 . The function Fxcn ) is non .

decreasing .

for see Enz

⇒ false ) , EFXCNZ )
s .

Right continuity .

t EIR
lim Fxlse ) =
Fx Ca ) a

Roat

for AER and 870

every ,

him ( Fx Cato ) -
Fxlal ) =
o

I -30

1¥
Decomposition Theorem .

If a function fun ) satisfies the

stated above then Fxfn )

properties ,
can

of functions
be represented as sum two

say Gxln ) and Hxfd ) as

Fxfd ) = Gxln ) t Hala )

where Gila ) is continuous and Hala ) is

a right continuous
step function with

jumps coinciding with those of film ) .

and

Hxc - o ) = o .
Ex ,

:
Fx

y
"

mixed

discrete
 decomposition if 9×61=0
In Lebesgue
the variable
C
identically ) ,
then random

x is called as discrete random variable .

If Halse )=o ( identically ) , then X is called

as continuous random variable .

The situation where neither 8 these functions

to call X
are identically equal zero ,
we

a
mixedandomvariable .

* &

If X is
random
discrete a
variable

, .
Pki ) =
fini ) = PCX -

- Ri ) > o A nie Rx

2 .
Z Ani ) = I

Ni

PCX Ski ) E f
3 .
Fxfni ) =
=

KERI

let ni and Rite belong to Rx such

between
that nitrite and no
point aid

to Rx
nite
belong .

Fx Geeta ) - Fxlni ) = FCK it ,

) =
PCX =
Rit , )
 Continuous random variable .

Hx then X
In
Lebesgue decomposition E Oo ,

is X a continuous random variable .

( s > o
)

Ex
PIX =
a) =
lim I Ints ) -
Fxln ) ]
8- so

=
O

False )
define false ) =
daz

T
for points)
"

exists except few

"

( always
a

in the above
example
at 4=0
,
fxco ) does

not exist .

density function
fan :
probability
Example
.

:÷÷÷x
Esta
-

CrF. p )
✓
,

uniform probability space

x : distance of the dart from the

center of the board C bull 's eye )

Let Faint
-

denote the CDF 8 X .

Fxlu ) - Prob Cx En )

tar c) uniformity principle

=

Tir

=
I
R2

°n2
Us O

Fyfe )
{
-

osu ER
£2
( I a > R " "

o
## R
The
pdf
false ) =
Ida fan )

f .
III.
:& r

( o n , R

{
false )
LI
okas R
=
,

° otherwise

Pla Exe b) =
Fxlb ) -
Fala )
b a

=
ffxlnldn -

ffxlnldn
-

a
N
-

= ftxlnldn
a

fpdT
 Let be continuous random variable with
x a

CDF Fx Csc) and false ) =

data False ) be its

probability density function C pdf ) .

Then the

of x is the set Rx EIR such

range
that Hut Rx fla ) > o .

Definition : A
density function or
probability

ffa
density function ) is a non .

negative
function such that

fffnldn
a

= I

The obviously ,
R

Fxln ) = ffxfnldn
→

CDF
Satisfies all the properties 8 .

b
fxcab ) ExCa )
Pla EXES )=
ffefnlda
-
=

Examples :
3

1) for what value k

,
foe ) = ka osx El

-
- o
otherwise

is a pdf ? ?
 § kn 'dn=i ⇒
k=4

¥2
'

fan ← pdf
fx

for YE C 0,1 )

:*
"
Fx CD =

§o4n3dn = y

Ii
.
.

EI fix Ossett
)={
Gall
-
se
)

o
otherwise

Kik
 of Variable formula
Change
.

Let x be a continuous random variable

with pdf false ) .

find the
deity of the

random variable Y= x ?

Let Gyles ) be the CDF of Y .

X
Let Fda ) be the CDF of .

Gly ) = PCY Ey ) for

any
real
no YER
pfx2Ey )
.

=p try ex 's Fs )

Gy Cy ) =
Fx Cry ) -

Fx C- Ty )
In order to find density 8 Y , say 91g) ,
we take derivative 8 Gyles ) .

%) =
day Gyles )

=
day ffxcry ) - Fxfry ) ]
x'
@ fiery
crude ) ]
fry
=

|9lD=÷[f×CB)+f×Gry# -
is > 0
Ex :

IT Exe R

{
o

false ) =

o otherwise

Find the density of Y=x2

Let gls ) be the density of y .

9197
ztyffx Cry ) tfxc B )
)
-
.
-

R2
[ 2¥ )
o Eye
Try
= o
t

|9ly7=tpz0ESER# = o otherwise
-
uniform

I Fina
?:

ii.
-
 Continuous uniform random Variable .

Let x be a continuous random variable with

the pdf defined as

fatal Ia ask Eb

f
=

o
otherwise

for two a b EIR such that aeb

any
-

Then x is said to follow uniform density

with parameter ad b .
XN Uca ,
b )
N

False )=
Sfx da =
I
b- a
aancb

-0

{ On
a ca
Fala ) =

a
deneb
-

b- a -
-

I N > b

¥÷÷¥
CDF

I -
EI Let XNUCO , , )

fxloe
{
I Osn El
)=
O otherwise

{
Beso
False )= O

R OGRE I

I M > I

Consider the transformation

Cl ) for X > o

¥
Y= X
log
-

Let
Gylg ) be the CDF 8 Y and gls )
be the pdf 8 Y .

Gyfy ) = PCY Ey ) Ty EIR

=p
f- ÷ logic - x ) Ey
)
=p [ toga -
x ) Z -
Xy ]
is
=p [ I -
x 3 e-
]
is
=P C x El -

e- ]
Gyly )= TY 970
-

I -
e
 gcg )= day Gyles )

]
t 's
II
-

e
-

day
-

|9CD=Ae⇒y# YEO

Xining
= O

th

parameter H2o) .

tease : Let X
Nexplx )
i
f
"
false ) de
'

=
n > o

o otherwise

§#
pdf
also
fxfx ) =
O

=L -
e- Mazo

For real s a y > o

any no ,
.

Mnt 's '

Fxfsety ) (I )
-

PCX nty ) I , e
-

>
-
=
=
-

Tenth in As
= e- =
e- e-
 dad
An -

pcx > sets ) = e- .

e

plx > mtg ) = PIX > a ) .

PCH > y )

PCX7s )
=
PIX ? n )

plx > y )

-) Cx

#mmlaspopertg
/ ) =P > a )
PIX > sets x > y

exponential density .

theorem : Let 4 be a differentiable function

which is strictly increasing
or
strictly
decreasing on an interval I .
Let PII )
'

denote the of of and O be the

range
inverse of Q on I . Let X be a

random variable
continuous
having density
fan ) such that false) f o on I .

Then 4=91×1 whose density is given by

Sly ) = flatly , ) I day 97931 DENI )

Roots Let Gly ) be the CDF 8 Y .

oily ) = p ( y Ey )

peek ) )
}

A
-
-

Ey is
p

= Plxsgtcy ,
,

Gly ) =
FILM )
⇒ gig )=
decreasing
Fx
'
Cutlet day 9453

gly7=fxCdcy)dqy
-
A is a function .

Gly ) = PC Y Ey )

41*3
PC Ey ) 9
}
=

pcsexzyqg
=
, ,

ftp.i-r-xcceky#9ts7=-fxfg7sDdqC94y
) )
-
.
EI Let X be a continuous random Variable

with density f .
let a
,bElR and beto .

define Y= atbx Then what is

Then .

the density of Y .

gb7= )
flaky
) 97971
) day

Cl In ) = At bn =]

94ns
Iba
a
⇒ =
= )

⇒ ÷
Y
91g )= f ( ,

seer

f
fake )
2£
OC
E =
,

O otherwise .

Y= I I
=

Ry
o aye
R

gly )
÷
f
) these
,b='z
a- o
=

( Hr )
Lj
=
2

RT
 919 )= 29 O EYE I

= O otherwise

Symmetric densities :
-

f is symmetric density if AnS=ff -

nd
)
txt IR

Ul -

a. a ) is symmetric .

A random Variable is symmetric if

its pdf fxln ) is a symmetric density .

EI .
If X is a
symmetric random variable

with CDF False) .

Then Fxlo ) =L .

-
R A

Fifa ) =
f fly )d3 =
f f C- g) dy
-

•
N

=
f faddy
2

= I -

False )

Fxf -
D= I -
Fxcx ) Ix EIR
Ex .

glass I - aero

'
Itn

? ?
Is gcn ) a
pdf
A
A

f tnzdn ul
"
tan it

sina.ie#-.anI
= =

is a density .
Cauchy density .

Since gcse ) -
gl -
a ) , Cauchy density is

symmetric

Ex e
-
n
42 a a see
gcse )
-

: =

MZ
C
f da
-

=
e
-

x
a
x

sik Tk
f
'

f dis
-

c =
.

dm e
e
-

N
-
N
 a A

44542
f
'
dy da
J
-

a e
=

-
-

a a

Polar coordinate substitution ,

IT
ik
do dr
or §
-

g
,
re

O O

& = ZIT

a = Tut

n % Fut
f e- dn=

-
a

Define
2/2
2

A Ln CD
L
-
-

QIN -

.
e
rut

clearly this Pln ) is a pdf .

This density

is called as standard normal density .

The random variable to this

corresponding
density X is called as standard normal
, ,

random variable .
 standard normal
density 4cal is

clearly symmetric .

#E
,
Pdf 41*1
- -

, It
Y I

a b

Rx = IR as alla ) > o Fa EIR

Pla Ex
Eb ) = Fx Cb ) -

Fxla )

let Ex denote the CDF of Std . normal

random variable .

Pla E X Eb ) = Ex Cb ) -

# la )

! else ) die -
-

⇐r*e"%dn

It seems that it is enough to know

Ex Ca) taek
 I
R

Ean ) -

-
f 914dm
To

0
Let a >

⇒¥¥¥#=s±
-#
=
f-
IIIa )
)
Etta
, -

oIxla )

0
a

Ext -
a) =
I -

Extra )

Evaluate Ex f- ) I Ex la ) ca )
-

a
-
=

ppl -
32×23 ) =
Ex Cs ) -
Exc 3) -

=
Exes ) -

[ I -

Extra ]
= 2 DIG ) - I

= 2 ( o
-

9986 ) -
I

PC - 3 2X C 3) = o
.

9973

PC -

I exe -

tu ) = Exc ta ) -
-
Ext I )
=
Edt ) -
Exc 'T )
= o -

o 928 & )
 Let be standard normal drandom variable
x a

Define Y=
puts X where r > o

find density f Y .

Let glg ) be the density .

EM
'
9197 =

I 9C )
¥1
"

, •
ay

#
e-
-

This Y is said to follow normal

with two parameters do ?

density µ

YN NLM ,
r2 )
I
" ¥12
91g ) =
I e
-
←
⑨ eyed

Tz

Probability computation
PC as Ye b ) =p ( as lutrxeb )
=p f a -

praise
b -

µ )

=p ( Fg ex -
HI )
=oI×(b¥ ) -

Extra )
 ,rY

x.n.IN
8 Nyu
gpdf

P ( In
-
36 a Y s
fit 's )

=p (
M
-
3T -

M L Y -
M a tutor
-

µ )
I Y
?
M 3€ )
(
- ,
-

=p

I
(
. n
. .

a
=p -3

= Ex I s ) - F -
I

=
O .

9 a
Gamma densities .

let x N N ( 0,62 )
Y=×2
Find the density 8 Y .

n 4202 co
fin
Iz
• an
-

) e
.

let gig ) be the density 8 Y .

915k¥ ( Ary ) -1 H -
Ry )
) so

Kok
91g ) 1-
-
b > o
=
e

TE
Nok :
g Cg ) is a Gamma density .

Gamma functions :
-

"
in

§ r¥
-

n
e
dm =

Gamma pdf :

n' the
rcn ; a
,H={f÷ ,
e- a

N
, o

Io
O

parameter f this density are a

, A .
 take
2=12 ,
D= z↳

EFtahonofarandmvanm.rs .
V . S .

/ -
discrete continuous

d. f
Rx

-
p
.
-
.

Countably
finite
infinite

f
p.m
.

.
-

be discrete random variable

Definition : Let x a

with Rx as its range

Ext
Ese
UGH
Pk -

as 4jovicdoendvertgheeu.mn )

*
Ex : let be a discrete uniform riv .
on

{ Hi
,
are
,
.
.
.

,
In ) I Rx

for nie Rx
PIX ni )= In
-
-

I O
-
W -

O n

EH ) -

.fm?ipcx--ni7--??gMiPlsei)--E7ni
.
 uniform discrete ECX ) is
f
riv
In
.

case

nothing but the A. M .

f Rx .

In
general , we can understand Efx ) as

Weighted average

E Xr Bernoalilp )

mulish I ←
mttasfuasiunm
.

ECX ) -
- o .

( I -

f) t I .

p =p

X Binomial Cn p)
EI N ,

" "
PIX a) ) for 4p ) a- on in
- -

Cnn
.
-

=
-
-
,

otherwise
= o

'T
EH ) -

¥ .
j ( ! ) pig
pgn .

j.nt-ji.in?jy.=nCII
notice : )
-

ECH=np
Ex :
XN Poisson IX )

PIX -_
a) = ee a - on ,2 ,
- - .

se !

O . W .

A
j
Efx ) -
E j e
j=o I !

= d

EI XN Geometric Cp )
"
PIX -_
a) = ( I -

f) p se -
- 0,42 ,
- . .

= o
otherwise
A

texts
Ejpc
j 's
, -

pi
'

Ejci
a
I
pj
'

= Pll -

p )
-

j=o
I
§!dqpa
= -

pcl -
p ) -
b)
 E and day because
Interchanging ;

The Aeries is absolutely summable C accept )

pj
Io
Ek )
-

Pll -

P) Cl
dgp
-
=

Efx ) = II
p

txanph
s

Let be discrete r r with the .f

× p.m
a
- .
.

÷→
{
se
fin ) -4,2 -
-
-

- ,
,
-

O otherwise .

clearly fin ) > o t NER

,

and
Eiffel -

a.I.cat ,
= fat -

att

-
-

C ⇒ HE
-

theft -

He -
-

=
I

ECX ) =
§ se - the ) -
-

§
" .

qTn+
, ,
=
£,
It ,

exist
,
,
does not .
 # A 8 discrete r r .

a
.

Expectation
Let 2=91×3 be a function of the

discrete random Variable X .

function
) alla ) PIX )
£
a
El 2) El 41×1
-
-

=
=

provided that the exists

expectation .

Ex , X a discrete r . r .
with
pmf

×-2-#÷
PKII.is/Yslysfys/y#
Efx ) EH 4
,
It 11

2
o

properties 8 expectation .

#
ret

bear
Let x and Y . v. s .
with finite

expectations .

1) for an DEIR such that PIX =L ) =/ ,

EH )= 2

2) for some constant CER ,

Efox ) -

-
CECX
)
 DECK t 7) =
Efx ) t Ely )
PIX 391 ) I Efx ) 3 ELY )
4) If = then .

moreover
,
Efx ) = ECY ) if d
only if Pk=Y)=l .

5) I ECX ) I E ECIXI )

Exercise :
Hypergeometric
Population Cr )

I I
I
Type I Type
or r
, r -

A sample of size n is drawn from

the
population .

Sn = number of objects 8 Type I in this

sample
Define : Xi : ith indicator random variable
ith
indicating whether athe object in the

I
sample is of Type
.

Sn =
A tha + - r
.

f- An i Efxi ) =
In .

Eon ) = n
Ir .
 Moments :
for a fixed r E { 1,2 ,
- - -

,
}
-

) ur PK a)
§
' -

Efx
-

These are called as raw moments f X .

8=1 is Efx ) also called as mean

q X .

'
air
@ PIX )
{
a
)
-

EIX
- -

-
a =

for some fixed AER .

Central moments 8 X .

Take a -_ o Central moments are

,

Same as raw moments .

take
In
particular

a=µ=E
for 18--2

ECX -

pet = ECX -
ECHR = variance I X .

root the variance

The positive square f
g x is called as standard deviation off .

EH -

-
µf=r2 ← notation
 82 Varix )
Interpretations #
8 .

measuring squared variability 8 x

around a

2
= Efx -

a)
2
Efx rt
Interned in
minimizing a) w
- .
.

a.

constant
( Approximating by
r v a
a . .

and the error in the approximation is

EH -
at ) .

)
'
ECXZ
'

Ek zaxta
al =
-

= Efx
'

) -
za EH ) t
AZ

diff .
w . rt -
a

-
2 Efx ) t 29=0

ECH=
⇒ a
µ
=
 Another f variance
interpretation .

Given a random Variable x and

a number a .GR .

CXat -
-

= [ Cx µ )
-
t c µ
-
a) 32 when
µ=Ek )

=
( x putt
-
I in -
at
+ 2K -

µ ) ( µ -
a
)
2 2
ECX -
a) = ECX -

gift Effi a)
-

+
2ECµ -
a) C
fad
)2= ECK
°

* a)
2
Varix )
- a
Yu -

Chebyshev 's inequality .

Let be random
x a non -

negative
variable with finite expectation .

t > o positive real number

any
.

Define a new random Variate Y thanx

if et
4=0 x

Y -

- t it x > t
 4=0 if x at

Y -
- t it x > t

Ect ) =
o .

PLY = o ) + Pk > t ) . t

ELY ) = t PCX > t )

Note X > Y

Efx ) 7 ELY ) = t pix > t )

/plx7HEEC#

270
⇒ -
ca
t
#

Chebyshev 's inequality .

@FtHuI.T
X be a rev with mean &
gu
.

variance r ? for real t >

any o
,

Consider ( X a) -

'

p
E
E¥M . .
from #
 function Cma F)
Moment generating

Mxltl =
Elek ) for a given
random variable X .

In case
8 a discrete r .
v . X with

p.m
-
f .

PIX =
ni ) = fini ) for Nif Rx
,

MxlH= Ecetx )

qetni pcx ni )
-
-
=

Ni

EI Xv Bernoulli I p)

pg
I et
I
et
.

MxHI=
.

,
-
+ p

MxlH= etfs + Ci -

p)

Ex : XN
Binomial Cn ,
f )

Et et
' ' " a
MxlH=
-

p em
" a
n -

Eoc : ) Ktp )
q
"

=
Gtpet ,
Ex :
X N Poisson IX )

die
Zeta
-

Mxltl = e

x=o It
A

Ceta
se
i
2
-

= e

se = o a !

= e
-
d get
-

consider XN Bernoulli f p)

Mxltl = (I -

p ) + pet
adz Matt ) = bet

XN Binomial Cn ,
b)
h
Mxltl =
-
p ) Cl
+ pet )

d
Malti
d't
,
 Ecetx ) )
= Ell ttxt
t÷'t ,
t÷x?t - . .

E(t¥
Ectx )
ECI ) )
t t
=
t
-
- .

?t EW )
tf
'
Milt ) = LA t
t EIHT t EH )
.

T - -
.

It It, EW )
'

Mxlt ) EH It Efx ) t
33.4-2
=

of - -
- -

Idf Malt ) I = EH )
t > O

!
d2
Mdt ) = Ekg
at ⇒

1) Mxlt
Malti is always defined
) =/
-

o
,
-

for t = o .

27 Matt ) as a FI of t should be defined

in a small interval around t -

- o .
 let -
x be a r . v .
with Mbf Mxltl .

find the Maf 8 the v.v .

axtb

for some a
,b C- IR a to
,
.

b
#)
Elek
t

Maxtb CH = )

EC eat ebt )
'
= .

= ebt Efeatx )

=ebtM×Cat )
-

Expectation of continuous rn
a
.

Let x be a continuous v.v . with density

C pdf ) false ) .
Then expectation f x is

defined as
,

false ) du if it exists
Efx ) =
f se .

-0
a

[ ECM exists it fhelfxlu ) due a

]
a
-
Ex : X NV la ,
b ) b
a

I
nefxlxldn f se
du

f
.

=
ELM =
b -
a

a
-
a

Tin
÷:{:
"

{
vca.si ; fin -
-

u E. ⇒ : =
¥

XN Gamma Cd A)
E
.

de in
fxla ) I N 0
-

>
n
= e
rid )

0
otherwise
I

§nf¥
' in
EH ) -

-
nd e
'

du

II ! II ;rY÷
"
a- e- an
?
-
-
-
-

,
 xd run )

|EH⇒
=
L,
- -

TIL ) Anti

X
-

SXN Gamma ( X ) d
E±
I
, ; > o

T "

fan )= de
-

noo

=
0 otherwise

Gamma I lid ) eypld ) d) o

;
-
-

ECX ) -
-
L
x

EI : let fxln ) denote the density of

Cauchy random variable

fxfn ) =
1- -
acres

Tofte )

Eun ) =
I
.IT#da--%7EadncauTydensiq=Inhjm
da
. !
have finite mean .

= ,
 Moments 8 a Cts .
r - ✓ '

with raw
f am
.

Efxm ) fxln ) du
.

c) =

moment .

→
•

4 Ek -
at =
f almfxlnildn
@ -
:
with central
→ moment .

M 1,2 ,

µj=
= -
- - .
a

variance f X = oh = Efx -
f @a) fxlnldn -

TN

where EG )
µ
-
-

EI xn gamma C a
,
x )

§
' t "

am nd
Efxm )
g÷,
-

= e du

tu
amtd
§
- '

da
X÷
-

=
e

Xd

rhythm
=

-
r Gtm )
-

|EHM= dHtDnHtm
-
 '

variance
f x = r2= EW ) -

LEK) )
-

nets -

E .

→
/r
EE In particular ,
if xnexpld )

Efxm ) =
MI
dm

Cx )
¥
var =

EI Easy exercise .

X NU la ,
lb]
' '

EIN ) & Ia da
b÷a
a
beta
=
= .

'
'
varix ) -

atazbtb -
Gatsby =

" I,

'
ECT )
( EH ) )
Advantage of symmetry of a r .
v .
x .

Let x be a
symmetric v.v .

⇒ x and -
x have same density .

for any integer m EM ,

'm
xm and have density
Ex) same
.

for m odd .

Xm and Xm have same density

⇒
-

ECXM ) = El -

Xm ) = Efxm )
⇒
-

⇒ Efxm ) =o

In
particular ,
mean 8 all symmetric
densities is zero .
( Ex .
N 10,62 ) ) .

functions CMGF )
Moment
generating .

Mxlt ) = Eletx ) it it exists .

EI X NN C µ ,
o
'
)
be -

M2 IF acne ,

fan )
g¥
- -

= e

tack put }
Ira Mpf
-
-
-
-
acnco
-

,
 a

¥2
2
Cn -

a)
tr I
-

Mx Its = f e
FEIT
e da

g n
µ
-
=

""'t
-
¥2
Jae frat da
e
math

It Er )
A
'

÷
-

=
eats ,
dy
→

Notice: to -

II .
=
- It"
" → %? et hay
"

mxiH-ett.IT
-

e
,

I #r*e¥
"

= eat
'
dy

L Nlr
't
,
r2 )

\
y
It
'

/ ett
'

Mx it , =

-
 × N N ( fu ,
oh )

E = mean of X =
dat Malti /
t -
- O

=§)
Iot
eat
-
'

E- O

④ etat Er 't
'
-
'

=
u + rat ) )
to

I
=µ) -

Evil =

Cat Mtk ) )
to

to 't
ett Katy
' -

eat
'

get
-

fda
+
=
µ day
to

Atµ
2
EW ) =

varix ) Efx
'
) -

LEAD
? Ttu
' '

gu
= -

fvarlxl-e.PT/
-
Exo Let * N
Nco ,
' )

My #
=et% - acts @

Define X =
fatty r > o

ut

Mxftl e' My Crt )

=
Mµ* ,
Ct ) =

= ett .
eMtYz
"kr2E2
= efutt
⇒ An NLM ,

In
general ,
if X NN ( mean
,
variance )
)tt @
asiana )t4z
et
mean
Mx ft ) =

Ex : X N N ( µ ,
or
)
Define : Y a
XII
r

if IX
-
=
+
-
n

b a

Maxtb It ) =
My CH = et4z
Example: xn gamma Cd ,
X)

§
" X "
Mxltl etn
f÷
-

= a e dm
,

' " t '

f÷ ,
"

§
-

=
a e da

=
Ad rld )
-

atf
-

ND -

( gift )
Malti . -
acted

EI Exponential density with parameter .

MxlH= I exp Id
) =
gamma Cl ,d ) .

X -
t
 Quartiles .

"

EIR is called 1st

A point 9 ,
as

quartile

P ( X E 9.) =
La .

A is called and
point MEIR as

median if
quartile or

)
PIX E m =
Iz
A called 3rd
point 93EUR ,
is as

quartile • if

PLXE 93 ) =
Z
Mode :

× a v.v .
with density Hln )

M fxla )
=
argmnax
 Let Binomial C 3 I )
E
XN , .

median & mode

compute .

" 3- "

33=0
(I)
on
Axl a) (L) ft )
a- ,
=

otherwise

Hl÷i¥¥#7
-

mean =
Zz

mode = I 2
,

Ill
mean .
' "

".
"
i
Ex . X N NC µ ,
or )
mean
1×7=14
median Cx ) = te

P CX Em ) = 0.5

⇒ pl XII ⇐ MII ) =
as

( MIIIE) -5
o
⇒ =

⇒
MII o
-
-

⇒ M -

µ = O

M gu
⇒ =

mode Cx ) -
-

tu

EI XN
Uniform ( a. b)
)
atzb
Mean IX =

median I x )= .

Fcm ) -0.5
MEL =
I ⇒
/m=7
 )
Ex : AN exp Cd .

Fcm ) -
.

Iz
im
I
-

e =
Lz
-

E± A
point is chosen randomly between

the interval C- 10,10 ] ( by uniform

probability

principle ) .
Let x be a random variable

defined in such a
way
that X

denotes the coordinate the chosen

of
point if the
point belongs
to f- 5,53 .

X takes value -

S if the points
belongs to C- to ,
-57 and X takes value

#
^
- '

5 ynnt belongs to L
-
I
.
-

, .

CDF 8 X .

Exit with p.d.fi

bet
f- be ) = !
-

Compute PCIE 1×112 )

 Random vectors
-

I p )
CR , ,

be discrete
let
r
Xi ,
Xz - - .

,
Xr
,

random variables or Cor ,

F , P )

I
Define dimensional

)
r
is an
-

1¥
;

IE
vector

Ilm
.

Huh
for an w fr

err

Xzlw )= Kz Xrlw ) -

ar
Kfw ) se
-

Suppose
- - .

;
=
, ,

f! )
-
-

Eir

( !! )
be rector
Definition : Let =
a

where Xi is a random variable on

Car ,F P ) For Rr
, .
every
 the set { wer ) ECIW ) =
E } EF .

Then X : R →
IRR is called as an

r -
dimensional random vector .

we are interested in Prob ( I -- I )

•

X. Cw ) discrete random

µµ )
let X=
-
be a

,
-
Vector .

Xrcw )

If I denotes the value assumed by

the random Vector I ,
then

{z ) finite
: PIE = I ¥ o
} is

or
countably infinite .

Definition : The discrete density ) discrete

of vector I
joint p.m . f. the random

is defined as :

f- C
Ry . . .

,
Ur ) = Prob = Ri ,Xz=N2 ,
- -

,Xr=xr)

)
where and
E-
Her ) E-
1¥
.
 In the vector notation

f II ) = PCI -
-
I ) t
If Rr
'

For a subset AER ,

PCE E
A) =
2 fire )
IEA
Definition 1

A function f is called as discrete

joint pm .
f .
if

IE Rt

ffzi
i FCI ) t vector
) z o
the r .
.

Range 8
fte ) }
←
finite
) E I
or
is
ii :
to
infinite denote the
countably . We

set 9,1373
elements f This as
,
.
- -

) I
iii
) ?
-

-
 s¥i7

In
E#
'
14 48 " 2

Tim
€16416
PCI I )
( I: )
-
-

Is
=P ( X
,
=
Hi ; Xz
-

- ku
)
= fine ,
Nz )
= fine )
table
Very easy
to verify that this

is a discrete joint pmf .

Prob (X )
3×2 t
Tf ft
=
,
the t

Prob I x
, = I ) = Pub ( Xi ' I
;Xz= I
)
T Prob ( X ,
= I ;Xz= 2)

+ Prob C X ,
-

-
I ; XE 3)
t Prob CX , =L 5×2=6 ) =
42
 Pnb ( X, =L ) =
I fcl ,
Rz ) =
Yz
M2

flesh ) Yz
Prob Cx )
=

Enz
-

, -
z =

false feel )
)
§
, = Nz
,

,
If

of
pm .f
A X
marginal .

Independent

Problx,=rDfrqeR×€
r.rs
i
-

Let be discrete
,xr
r
a ,
. - .

,
x ,

random variables with f

p.m
.
.

fi ,
fr -
- .

fr respectively The random

, ,
.

Variables Xi - - -
Xr are called as
, ,

mutually independent if their joint

.f f is given by
p.m
.

flank . .
.
Rr ) = floe ) ,
.
fake ) .
- .
.fr/ser )
, ,
 Nothin :

f Hi
,
Nz ,
- -

, ng )

a Prob ( X ,
-_ se , g A -
- Nz ; . .
.

€
Define A ,
Er sit .
A ,
-
( x xD ,
-_

11

fish ,
Aifw
: Xilw ) = a ,
}
.
. .

,ar

;Xr=Rr)*
,

Prob I Achfzd a Ar ) independence

y
- - .

torrents
-

-
Prob I A ) . Prob f An ) .
.
pwbfa ,

=
Prob IX. = 'Re ) .
. -

Prob I X
#
Rr )

= film ) # , oh ) . . .
-
frfner )

the '
14
-
 Let and be two
Ex
:
X
, Xe independent
random variables each with
geometric
distribution with
parameter p .

find the distribution min Cx #I, .

(
Xp
Prob min I
,
Xz ) 37 )

Prob I X, 3 Z
j
X 232 ) independence
=

2
844×2
=
Prob ( Xi 37 ) . Prob ( X , > z )
? t
-
-
-

f p)
.

(I -

f)
't
=
(I p ,
-

Minix , ,xz ) N
geometric (
I -

Ci -

pp )
 Discrete .

Bum of independent random variables

two independent
s
Let xey be r r
. .
.

Let be the distinct

oh ,
an . - . .

values taken by X .

Interested in the event

{ Xt Y = a } t z

11

U { X -

- ni
,
4=7 -
ni } ←
quote
his
, union is

disjoint .

PCXTY -

- z
) =p ( ?
EX -
Ni ,
't 2- -
ri
})
4=7 )
?
= PIERI ,
-

ri

PIX ni ) P 14=2 ni )
?
-
-
= -

-
fxtyCH-ffxlalfylz.se#
 discrete
Expectation
-
.

)
be
( ¥;
let
a
given

Larandom
I =

f
vector with
joint p.m
.
.

( sur )
fxy Us
- - .
-

. .
.

,
× ,

be function of
Let him
,
- - -

per ) a

A Rz - -
Rr .

, , ,

Echlxn ,
- - -

,xr ) ) =

[ hlay . -
-

Mr )fx ,
,×z ,
. .
.
,xd% - -
-

Mr )

Mn - -

,Nr
,

For r=2 hlse ,

,
da ) = kit Kz
,

)f×
@ ,×zfh,k )
?
the
Ecxitxz ) =

Ey ,
 r =2 htt , ,K ) = di
,

fxyxz
C seism
)
§ §
Ela ) =
see

?
m
)?fx,,xzlmie

/Ek,)=Emfx,Ca
In general ,
take HIM , - - -

per )
-
-
Ri

for # er .

' Eking -

t
 m

and Y
Let X be
independent
discrete random variables .
with
joint
pmf f*,y
His ) .

fx Ca 'D
§
§
as
) y ,
ECXY = ,

fxln ) fylg ,
§ §
My
.

=
Ea
infxln )
Eg
yfyly )

|ECXY)=EIx7E
-

4211¥ kid
ECIYCX ) 4am ) ) ? § 4. In )
=
,

ft , Gettin ) )fy/y )
=
NE 54219

=
Else HM ,
Ethel 'll )
 x dy are
independent discrete r.rs . with

Joint Pmt fx ,
Ca , , ,
M G. f .
.
8 ×+ y

Efetcxtyyy
)
Mxty It ) =

= Efetxtty ,

Eletx et " ) independence

=
2
.

=
Efetx ) .
Et et 's
)

Mxtyft ) =
Mxlt ) Mylt )

Generalizing this result

,

x.
gxz ,
-
-
-

,
Xn are
mutually independent
discrete r .
v . s .

ME.a.CH
=

II ,
Malti

( p) for i
EI Let Xin Bernoulli -44 .
-

,n

be iid dependent &

CinIdentically
Ihistnrbuted)
n

ya Exi
i =
s
 My It ) =
MEECH
n

= IT
Malt )
i Is

pet )
n

f I p t
-

= IT

i = '

's
=
(I -
p + pet )

Binomial l n
, p) = sum of in independent
Bernoulli It )

IMP observation
-

x & Y are independent

⇒ ECM ) = EH ) Ely ) already

-
proved .

Q :
Is the Converse the ? ?
NI .

EE let His ) be a discrete r . v . with

range Rex ={ ( o
, l ) C 0 -17 ,
11,0 ) ,
C -

1,01 }
, ,
, ,
with each outcome equally likely .
 for His ) t
Rex
p Cse
, y ) =
tf y ,,

-
- o otherwise

Efx
) =
o

Ef 'D = o

Now
,
XY = O ⇒ Efxy ) = o

Observe that

'
PIX = o
) = 12

PM = o
) = 42

PIX ) Plx =D Ply )

of
= o
; Y
o
-

-
= o =

=
Yz .
Y 2

114
-

Remember :
-

Ecxty ) =
Efx ) t ECT )

E fax ) = a Efx )

EH )
Efxtc ) =
t c
 Sum of variances :

let Xd Y be discrete random

variables .

'

varcxty ) =
E ( Xt 't -
Efxty ) )
2
= Elxty -
EH ) -

Ely ) )

CK
Z
= E -
EH ) ) L t Y -
Ely ) ))

?
= ECX -
ECXDZ t ELY -
Ely ) )

+
2E(X-ELxDCY-t
= Cov CX , -1 )

early ) -12Wh 'D

varcxty ) = varix ) t ,

Note :
covcx ,y)=E( x - EG ) ) (Y - EH ) )
= E ( XY -
XECY ) -
YEH )

+ ECXIECY ) )
=
ECXY ) - Ek ) ECT ) -
ECXIECY)
TECH EC 'D
|covCx,y)=EECx)E
Corday : If Xd Y are
independent ,
then

an LX
, -17=0 .

Converse is NIT tone .

,y)=o
&
Y are
independent
⇒
WH x .

If Xd Y are independent r .
v. s .

Varkey ) = Var t var C Y)

-

'
Varcax ) = Efax - Efax ) )
= Efax - at 72

lvarlaxj-a.ua#7-varCc
'
= of Efx -
ECM )

where cis constant

)
.

ECc c) Z=
'
2
-
Eat ) = Efc -
Eko ) = O
 let Xi
,
xz
,
-
-

,
Xn be independent
random variables with variances

92,65
'
- - -

rn
, , .

var ( X , t Xzt - - -

t An )
= var CX , ) tvarcxz ) t -
-

f- Varlxn )

= of trzt - -
-

t Tf

let Xi xz - - -

, Xn be iid with
, ,

mean and variance r ?

µ
( 't× ECE )
×
) El
ECI E
) tr t
-

= =

=
In Efx , )t . -
. -

its Efxn )

In pet t
tape
.
-
.

= .

|ECE)- -
 varix ) -
-
Var
( ×t--n )

=
vartan I t - .

t
var
IF )
-
-

ha [ E. varcxii
)
=
not
an

µartx7=II -

Correlation Coefficient

Let Xyy be two discrete r . v .

I .
Then

correlation coefficient
=
ex , Y ) C tho)
is defined as :

ecxiy
corCx=
e
en , a
-
- -
-

TETANY ,

= Corexit )

SDLX ) -
. SDLY )

Aly are
independent ⇒ ex , = o
 Schwartz
theorem :
inequality .

let xd Y be random variables with

finite second order moments .

?
)
?

[ Ely
'

ECXY ) ) E Efx )

Furthermore holds when

, equality
P(Y=o ) =/ or

P ( X
-

-
ay
)= I for some at IR .

Proof :

Easy to see Hat when

ply - 07=1 or PK -
-
ay ) =/
,

equality holds .

In order to
prove inequality ,

for any A EIR

)
MECH
' ? '

Of Efx -
XY ) = X Ely ) -

+ ECT )
 Since the above is
expression
a

A and Ely
?
)
quadratic in ? ⑥
,

the minimum value of this quadratic

is achieved at
expression

) )
'
2h 2E C
XY O
Ely -
=

1×1
⇒ a =

ECYZ )
The minimum value at the
point
ECXY ) is given by
⇒ ,

f ) "
E "
E 'M -2

t ECM ) 30

"
CELXY) )
⇒ EM
-

+
> o

Et
€sEH
-

⇒ a
Importance of Schwartz
inequality .

Apply this Schwartz

inequality
to two random variables

X - EH ) and Y - ECY )

"

L x -
Efx
④ ) )C y -
Ely )) )
?
E ECX - EG ) )2 ELY - EH ) )
2

⇒ @
Cx
, > ) ) E varix ) nearly )

⇒ ?
e = @cx, ⇐ I

,
Var vary )

⇒ lexyl E I

or -

I sexy El

/eHi)=Ii⇐plx=a' -
Reed :
If X is a non -

negative v.v .

with finite for -270

expectation then

HEEtfs
pix >

Chebyshev inequality ,

p ( Ix - >
put
t ) ⇐
If
-

Let Xi Xz -
- -

Xn be i. i .
d . or . u -
s
, , ,
.

Let
µ
- EH , ) got = varix ,
)
E
( SI ) =
Efx
n
) = la

) I
( SI) (
var x
= var =

Chebyshev inequality ⇒

PCIE -
at ⇐
rarefies =
 PCIE -

at > as
) ←
if
In particular ,

high Pfl In pet >

8) o
=
-

-
WLLN
Weak law
of large
numbers
-

Joint Continuous random variables

i

Continuous r.ir s
X and Y are
.
.

on the same
probability space .

Fla , y) = Prob ( X Ese

, YEYJ Joiner
Loo
-
A cap

Rectangle R -
-

{@,y) Iaea Eb
,

f )
•
caged

* .
 PLAY ) ER
)
=p I acxEb
,
c
a YE d)
= Fcb ,
d) -
Fla ,
d ) - Fcb ,c )

+
Fla ,

Marginal distributions

marginal
Fxla ) =
PCX Ese
) =
floe ,
a
)
CDF 8
= him
year
fluid &

Fy (g) =
PIYEY ) =
FCA 's ) marginal
as F
him
= Fla
,
'D 8 Y
naw
.

If there exists a
nonnegative f?

fin , y) over R2 s .
t .

Flu ,y)=
!↳ flu ,v)dvdu ,

then fin is called joint Pdf 8 H Y )

,y )
, .
 PIG7) ,
f A) =
f) finished ,
A

obviously ,
•
A

f) fin ,
y ) dndy =L

a -
o

and I floe ,y)= Hai )

009

Exi Let Gil ) denote the random

vector representing the coordinates

of randomly chosen point in the

center at lo , d radius
circle with o
)
R .
If
fer dey ) E circle
fin ,y)= C ,

otherwise
=o
.

Find C .
 E circle
His )
flu ,y ) =

¥2 of tg2sR

otherwise
a 0

dad ,
PCH ) C- A) =
Sf flap )
A

=
Areaoftt
IT R2

Else ) =
Flat ) =
lying,
Floe ,y )
A

false ) =

fffcx
-

a
,y ) dy
 A

false) =
ffxyla
-
a
, g) dy Marginal

pdf 8 X

dy
grantype
=

rain
.
.

Renner
fxln )= 2Th -

?
ITR

otherwise .

= o

Marginal
fylg )= ffxyloe ,y)dn
-
pdf 8
o
y .

)=2R
fyyly TRI
- Reyer

otherwise .

- o
 X and called
independence : Y are

if and only if
independent r.rs .

fxylr , >) = far ) fyly )

C
joint pdf is the product of marginal

pdfs )

In this case Xd Y are

NOI

independent
.

Easy way
to generate examples :

let x & Y be two independent its .

r . v. s .
with pdfs Acd ) & fab )

resp .

fxyla , y) = Aloofly )

fxycn y ) ,
70 try

SS Aca )fzly7dndy
{
Syfxycnildudy
=

finds Sfz dy
= S
 Ex :

x n Nco ,
i )
y or NCO ,
I )
Asseco
Mz
-

data ) =

tf e

- 542 cyan
Ugly )
a

LE
-

= e

fxyln g) , = 4×67919 )

449912
It
-

=
e -
Dogged

Let X and Y have

Example :
joint
density
in - ay
-15312 → arisen
foe ,y)= Ce

Find C ? ?
 53/2
'
Cn -
ng t
-

floe y ) =
Ce
co
,
→ easy

= o otherwise .

Marginal 8 X .

flu g) dy
false ) =

-
f ,

+54/2
•

Cathy
f dy
-

= e
c

- a

44735/43/2
•

fly
-

of
-

=
e dy
-

348
a
-@4512
f
.

dy
= c e- e

a
-

→
§
3h48 e-
2

der
= c e-
-
a

3h48
false ) = Fut c e- ⇒ C=T÷
 Distributions of sums
&
quotients .

fix ,y )
Let x dy beats .
r .
v. s .
with
joint pdf
Define 2=91×51 )

for a fixed a ER we are interested

,

in the event E 2.Ez }

By Az f R2 ,
define

Az -
-
{ can ) I olla , y ) E 't
)
Fzcz ) =
P ( 2 Ez )
= Plex ,Y ) E Az )
fine y ) da
dy
Saf
= ,

Distribution of sums ( elk ,

Y ) -
-
Xty )
Az -
-
{ he ,y ) )
ntyez }

¥¥k⇒
 Fztr )= f) fin ,
g) du dy

Az
"

If Axis ) dy ) da

Substitute y
-

. u -
n in the inner integral

Fdr ) -

Iff ! fin ,
u -

a) du
)dn

Fztg )
Ituano
=

!
Thus given by
the pdf f xty is

fftyf =

§ ,
fix ,
2- -

An ) doe
- NECA
If xd Y are
independent
A

fxtyl A =

f fan ) fylz -
a) da
-

&
→ LELA

If xdY are non negative independent r r s

- -

- .

fxty (8) =
f!f×la7fylZ -
a) da ocotea

elsewhere
= O

EI A Ynexpld ) are
independent .

tu
fxln )= Ae
-

x o
>

TO O . W '

fyly ) deity
920=0
.

O .
W -

0=0
2-

fxty (3) =
I false )fy Ca -
a) dr 2-70

O .
w .
 for 2-7,0
2-
he
Mt - n )
fxty CES du
-

f
-

= Ae de

O 2-

= Me
- dit
/ dm

ez
=
d2 At -

At
d2ze
-

fxty 177 2-70

-
-

O W
0
.
= .

This xty ~ gamma C 2. X )

EI X. Y are iid No ,
i )

fxty =
??

f false ) fy Ct -
a) da

O
false ) fyf 2- - n ) takes only Values

Zero or L

) takes value I when

fun ) fylz -
a

I
o En El
and o E Z -
as .

If o Ets I ,
then the
integrand
has value I on the set Osu Et .

f×+y( 2-7 = 2- OLE El

If ie Z I 2
,
then the integrand
has value I or
the set

2- -

I E RE I

fxty ( 2) =
2- Z i CZ S2 .

Thus
,

Anca
.

{ .
÷¥¥
÷ .

O -272
 to
EI let xrrldi X ) and Your Had )
,

be independent r . r .
s .
Then

Xty n r ( data
,
A)

ah
'
Me
-

Xd
film )
' '

= e a > o

f ( di )

ah
AY
'
ugh
-

fy (g)
-

=
e y > o

f C Az )
for 370 2-
,
L2
jet
iz
I
Li
fftyl # Je
-

Egg
se

, & '
-

O
I z -
a) doe
 ⇒ xty N RCL , tar X )
,

Ex : let x. Y be independent r . v - s .

with X NN ( Me ,
-92 )

You N ( M2 ,

Then what is the

pdf 8 xty ??

I Another approach is MGF )

Mxty Ct ) = Efetcxt "

)
= Eletx et ) ' '
since xd
c)
.

Y are

= Ecetx ) Ecetyy indep .

=
Mxlt ) Mylt )
Mitt rith µzttoEtYz
=
e .

Hitter)tt@Ttr2)tYz
= e

⇒ fty N N ( Hitter , offset)

 If X & 7 are cts .
r .
v. s .
with joint
pdf fx , ,
la , y ) .

Ecce IX. y ) )

dy
If
du
=

91ns ) face ,
g)

If x &
Y are
indef .

El 41K ,
'll ) =
If ya , y )f×fyfyly7dndg
If xey are
indep .
and Play ) -4,1*14214)
-

El 41×1417=1 f y ,
ex , Udy ) film ) fylyldndy

fdiklfxlnldufq.ly#1y7dy--Efd.CxDEf9dY77

[ Justification 8 Mbf result we have

used on the previous page ]

 Generalize the results .

iid for i
I Xi Nexplx ) are -43 on
-
- -
.
.

E.
Xi n
gamma ( nad )

①
Xin gamma ( hi ,
d ) are independent
for 5=1,3 - - -

in

E. Xin gama CE.fi ,

i )

3 . Xin normal I Mi ,
oil ) are
independent

fr I -
1,2 , .

r.sn?!XinaormalfE..Mi??ri2
)
 Distribution on
- quotients
Let Xd Y be two IT r . r . s with

joint pdf fxy In , , ) or the , y ) .

what is density for 2 - Ylx .

Az -
{ His ) I Hu Ez
}
If seco ,
then The E Z ⇐ 9322

Az = { Ca ,y ) ) seco and y > set

}

{
}
lay ) ) a > o
y Eat
u ,

Fxlx CH =
ff fin ,y)dadg
Az

! Ife!Hn,y7dy ) da

*
f://Imiddyf.dk
 Substitute I
-

- nu ( dy I ndu ) in the

inner integral

kyxth-%ff-affa.nu) ) do da

§ If !n ).nu)
fin ,
du da
+

!
I do
finds) -

fylxtn ) -

flat→
fin ,
are )dn

For Xd Y ,
non
-

negatiu and imdep .

Fy I I
In false ) fy ( ) da
z = nu
,×
022-6
 EE Let xd Y
be
independent v.v .

and ride d )
with densities run , ,x ) ,

resp .

22 -
I

rlditdz )
fyy× (E)
2-
Prove : =

tighter
-

447422 ) ft

Oct ca

= 0 elsewhere

'
al in
-

' '
Xd
Record f.
-

:
Cal = e a > o

DH
Ah $4
'
o
-

y >
fyly )
-

=
e
-

ride )

fylxlzt.mg?jtYq,zh-.Jahtaz-ie-seicztifn=pditd2zdz-
I
rfditdz )
-

Hd ⇐ti ) gditdz
, )rHz )
Ex Let be
: Xd Y independent
N I 0,02 ) r . v. s .

find the density f 421×2 .

'
x2
, y N
rft , Ira )

fy4×2
1-)
OCHO
C z ) =

ttfztl ft

= o otherwise

Homework ,

x & y are
independent Nco ,o2 ) .

v.v -
S .

find the density f 41×2 .

 Conditional densities .

discrete vector
Let 4,7 ) be r .

a .

Conditional
probability
plY=y)X= Pk )
) a y y
-
-
=
-

a
-

,
-

PIX = a
)

= fley )
fxcn )

where find ) is joint pmfar CAY )

and fxcn ) is the marginal pmt
& X .

Definition : Let X and Y be continuous

r .

v. s .
with joint pdf fin ,y ) .
Then

the given X
conditional density 8 Y ,

denoted as fylxlyln ) is defined as

fylxfylse ) = find ocfxtaca

f×Cn )
otherwise
= O
 Pla EYE b) *=n )
b
=

ffylx Lyla ) dy
a

Also ,
observe

If x dy are independent ,
fylxlyln ) = fylg )

Eixample
?Mt5)/z
:
( n

e-
1¥
ffln,y)=fxla)fyl×(Yl#)
An ,y7=
-
Den ,ycD

X n N to, %)
Cn ?Mty%
IIe
-

fishy )
fylxfylx ) =

¥
=

,
ITI ,g E

CY 725/2
¥
- -

=
e
,
 Thus fylx I Ma ) is NCE ,
i )

Prob ( OEYEZIX
-

)
o =
Iole ) -

Elo )
where Else ) is CDF 8 NIV ) .

Prob ( OEYE 2/11=2 ) = 20511 ) - I

EI let x N Ufo , ) and the

v.v -

Y NU lo ,x ]
find joint density 8 X. Y and

marginal density 8 Y .

{
ocsecl
fxln ) =
I

o elsewhere

{ Odyssean
fylxlylse ) =
Ya
elsewhere
o

fee ,y)= false ) -41×196 )

=

{
he
O
oeyc④ elsewhere
 A

fly )= f fin ,y7dn

-

&

=
f Indu
y

fyl 's) = -
logy ocyc I

=
0 elsewhere

Baye 's rule :

-

fxiy Cab ) = find

f y ID

= fxla ) fyixfyln )
-

)fykGlnTdn
Ikk
 Bivariate Normal density .

Random vector Cx , xn ) is said to follow

,

bivariate normal density if its joint pdf

is given by

Mpf f ( ITT
"
Haier ) =

asf 'Xn÷ ) -1
I
'
51 )
-
N CR , CD

LA
- A < Nz

The range
Probability computation :
8 parameters
P
( as Exists , ,
are EXzEbz ) →
Cpe ,
co

! Ian ,
. ;D .
.am/-:Y:::..
-
I sect
The
marginal densities
A

fxfui ) =
f fees ,
see
) does
→

e-
" C
' '
I
frat
-

•
-
ACK ,
c

Marginal density 8 X , is NIM , -92)

A

floe da
falser ) =

J ,
,
Nz ) ,

1K£42
-

"
e-

NER IN
- z

'

marginal density 8 Xz is Ncaa ,q )

ECX , )= Mi ,
Varcx , )=q2
Ekd =
plz ,

°
[email protected]
's

→ -

←
Efx ,
-

(guy
Xz Ms )
-
Then f= Corld
51.62

In independence uncorrelated ness

general ,
⇒

and tone
converse is NIT .

In case of bivariate normal density if

,

X. dxz are uncorrelated Cero ) ⇒

x ,
d Xz are
independent .

C =

jointof
product marginals )

the conditional densities .

fx = flair )
, 1×2
fxzfnz)

=q¥fzJ9p ftp.T.gy-fai-liuitekk#
u3Y
-
ffxzlx false
,
, ) =ffa
faces )

Frits In -

fµaskk#
-
ca ai ,
-

-
))
Thus conditional densities
,

Xzlx ,
and X , 142 are
both

normal densities .

xzlxi-n.vn/Mztef4q)Hi-MD,94I-5DXilxz--aaNNlMt9(9/r)Cdz
929-59
-

th ) , )
Ex :

( xn ,xz ) is a bivariate normal r .

v .

with parameters
0 2
Mz
= 1100 92 O 02
µ,
.
.

= = ,
, ,

q2 = 525 ,
l = o -

Compute Efxzln ) , ⇐ ) ( see

Matelski -

)
Mi

= 1100+0-9 ( F÷Lz
)
( N ,
- O .

2)

= 145.8N ,
+1070 -

84

Eckel 91=1 ) = 145.811 ) +1070.84 = 1816.64

Piz 7108014--1 )
varcxzfa , -4 ) = 0241-57=5254 -
o .

8D
= 99.75
Note that Y= Xzlk ,
= I is a

Normal and
density with mean
My
variance t
} given by

1216.64
My =

say = 99.75

P 1×271080/14=1 ) =
ply 31080
)
=p z
'D
108%41

⇒
1216.64
( 231080
-

=p

=p (23-13.6)
= I -
OI f- 13.6 )
=
Tell 3.61
I I
 of variables formula
Change
.

let Xi xz - -
X be random variables
, , , ,

with
joint pdf fish ,
- -

→
rn )

it
Define Yi
Eaijxj
- - r
n
= , ,

j= ,

A- .
Cai; )

!÷::]
=/ agin
: i unclean ,

t.at?xnl
If IAI to B= At .

(I;)
-

Bff! )
×i= &
.bijYj
I =L . - -

in
,

)
1- they
fly ,
,
.
-

,
Yn = .
. -

gun )
Idetl A) I
 non -
line as transformation case .

Yi = 9 ,
CA ,
- - .

, Xn ) i -

-
I
,
- - -

,
n

!
scan
8¥
.
- -

. -
en )

=/
,
, ,

÷ .
. . . .

floes
1-
fly ,
.
- -

syn ) = ,
- -
-

,
an
)
1J (see ; -

in ) I
a

Let A Xn be lid exp Id )

'

Examine :
,
. - -

,
.

Yi = X
,
t - - -

a Xi I
-
-

1,2 ,
- -

,
n

Y , = X I

÷:*
:÷÷ .
:÷D
⇒ Joint density 8 Yi 's is same as

density f 's this

joint in case
Xi
.
 Sunny :

discrete cts
joint r . v. s .

joint
I
distribution
joint pmf
marginals joint pdf
conditionals marginals
independence
conditional
independence
moments -
-

( Bivariate
( mean
,
Variance
,
median
) normal )
mgfs
Conditional expect 'm & Variance .

Covariance d correlation

transformation,

a.)
distributions of sums
"
-

" "

.
as
.si
:
:#
-

A ,
- - -

,
Xn ↳ X, f- - - -

t Xn

↳ A Ka

Rn Rn
Yi -
-
Bi Hi ,
-
- -

,
xn ) ;
i -
1,2 ,
- -
.

,
n
) >
 t
D standard
Sampling's
normal .
:

Nlo , , )
1¥
-

→ I distribution with n degrees 8

freedom .
Chi -

squared :
I )

Let x. .
- -

, xn be iid Nco ,
)
,

×i+---n unicorn )
"
=
4t-; Yarn

/
In "I I
Alon)
-

particular

II
In won )

nrftz I
)
for its
⇒ ,
 From the
additivity property 8
r ( di
,
P) ,

we know then
,

II. t Int
'
- -
-
t
II Nrc ? 's ) .

if ? , t ) is I density with n

degrees of freedom .

Notice : In is sum of squares f

standard normal
'

independent
'
n r -
v .
s .

( n : dot C
degrees of freedom ) )
Yi

Ym
Them : Let
,
Yz - -
be
, ,

independent I random variables

with dots ki
,
- -
-

, km respectively .

Then Y, th t t
Ym that
-

~
-
-

- - -

ekm

-
 Iki
ratio of i densities 'd
:}
?!
III
I
In
as
FCK he )
Let Y, ~
,

Yzlkz
Yz N ¥ K2

and Y, & Yz are

independent .

Then the random variable defined

the ratio is called

by

an F distributed random variable

with C kp.kz
) dots .

Flbkz )
Pastime h I =
: =
,
.

FCI kz )
soluaveofstdrnormal
ofstd normals
=
,

( of
)
Sam squares
.

divided by dof
 t -
distribution .

Let X be std .
normal v.v .
and Y

212 V with dof and

be r n
.

a
.

they are
independent of each other .

XNN ( 0,1 )
and xd Y are indep .

you 2x n

Then X is said to follow

En
t -
distribution with n dot .

× ~
tn

dearly ,
tf = FCI ,n )

it . w . Desire pdf of tn .
 Central limit theorem :

Let be iid with

Xi
,
Xz
,
-
-
-

pi
and variance r ?

To study the distribution of II. Xi -

Sn

Notice E Csn )
ripe
: -

Varon ) =
not

Suppose X , has density f then

,

fin ) if Xi 's
fqln ) Efsn ( y) y
-

=
,
are
y
a
discrete

fatal =
ffsn .
Cy ) fix -
g) dy if
are
xi 's

cts .

not : si -

sn-a.es#,=snr-nfInNl
0,1 )
 Central Limit Theorem C CLT )
'

'

Let iid with mean

µ
Xi ,
Xz
,
- -
-

be

and Variance r ? Let Sn - Xie - - -

txn

Then

him . pfsnjf-M.sn/- Else )

-
Assed A

where Else ) = CDF of nco , 1)

that )
Note
pfsnr-nfy-f.se/=Fsnahe
=
CDF f Sf
CLT status

In )= Else ) anew

nlim Fs
-

,
;

Fs : Ca ) n ,
•

I
as
- →

Else ,
 Observe

risner oIf÷÷)
:

EGILL )
-

*
) a -
-

rlsz e
III )
Generally for n 325
,
these approximation
are
really very good .

Example :

for i 32
nexpli )
-
n
Let Xi
-
-
.

, ,

Sn = X , t - - -

f- Xn

Prob I Sn En ) E OI
( r ) - Nemea

life bulb after it is

Example: Suppose a

distribution with
installed follows exponential
AS
soon as the bulb runs out
mean = to
days .
,

another bulb with same characteristic is installed .

What is the probability that 50 bulbs

? ?
are
required ion one yeas
Xi = life of it bulb after it is

installed exp ( to )
X Sgoz
,t
- -
-

-1×50

of interest Elsa )
The
probability .

µ
Plsso £365
) I E ( 36535g )

Tarif
= OI ( - 1.91
)
= O . 028
-

€:
 )
t
Pkn ) E
(

III
're #
a

let fsn be
pdf f Sn then
,

YENI
tsn a
¥9 ( ) area
-

OI : CDF f Non )
'

y : Te pdt if nco ,
, )

Ex ,
Xin eypli ) for i=b2 is
- -

iid

Sn X t Xzt it Xn
-
-

= ,

false ) ¥4177 )
aereo
I
-

Approximation for discrete r.rs .

false ) = Prob Csn = n ) 70

then a is a
possible value
f Sn .
 is Value
•
If a a
possible f X
,
then

Ae is an
integer .

'

If value of Xi ,
'
• a EIR
is a
possible
then the set

{ n -

at se is possible value
9×3
has god I

Ruling out v. v. s .
PIX ,
-_
1) =P IX. =3
)=I
Example :
-

let X N Bernoulilp ) perp It pls -

p)
, , ,

Xi ,
Xz
,
- - -

,
Xn are iid Bernoulli I p) .

Sn =
A t -
-
t Xn N Binomial f n
,
P )
n

pj
pay
-

tsnln ) -
-

(a) -

=¥ .sk:1#.D
÷:::::I÷i→f:÷
Prob Isner )
i¥ = E
( "Yz¥-) ers

EI Sn N Bin C p) n= 25 p o 6
-

n
.
-

.
,
,

pfsn 713 ) =
i -
Pkn El
a
El "%j%)
-
=
 statistics

Eskn
Statistical inference C and testing
of hypotheses )

Let Xr be random
Xi , ,
-
- -

, xn
a

sample from fol .

) and Fol .

where fol ) is the d f and Fo

p
-

. .
.

is the C- D.F .
of Xi 's .

( Xi
,
.
- -

,
xn a random sample means

Xi Xr -
- -

Xn are i. i . d . random
, , ,

variables with density fol ) .

.
)
that
Nole the
joint p
. d. f .

f Xi , - -

Hn

the is given by

flag .
-
.

,
sent II folk )
Ex : metal
X be random for
)
a
Let Xi -
-
-

, ,o
,
statical
sample from Bernoulli I p)
properties

do observe ??
What we

:*
X
,
= I
,
Xz =L ,
43=0
,
Xy =L ,
As =
0,4=0 ,

.
is

÷÷:÷::÷÷÷:÷
Feats available from repetitions of the

random experiment .

' '

Question : Can you guess p from the

evidence f date ? ?

obvious choice for f denoted

guess to
,

asf ,
was

I =
Ini
n
 note :

/I=Eff
I is a random Variable and

i
Eni
is a realization 85 .

for different set 8 repetitions of the

would
random expt .

,
one get
different values 8 I -

Effi .

The variations in the values 8 I

can be understood from the density
J I .

.
.
.
In this particulars case
,

X. N Binomial ( n P)
+ Xzt . .
.
+ Xn ,

"
an then notice
,

ECI ) =p =

E(×it)=n1 n

Vance ) =
roof ¥ + . -
ne
ftp.npltb )

rare )
Hi -
b )
=
→
Few definitions .

A
,
Xz
,
- -
.

,
Xn a random sample
from fol .

distributions
Det . I family of

F =
{ fol .

) I to is pdt 8 x

and at
}
e.g .

An
Let Xi - . .
. be a random
, ,

sample from Binomial C n

, p )

.tt 'D
}
"

{ (2)
B new
⇐ Beaten I

vector
Here g= Cn ,p ) : parameter

⑦ = IN x [ on ] :
parameter
bn
space

¥¥**¥ ← parameter space .

Example 2 : Let Xi Xz - - -
Xa be a
, , ,

random sample from N Liu

,
T ) .

N=£÷i±" }
vector tarmacs
A

ftp.E?aEEnEt
space

in the
Note :
Any vector demoted as A

"
"

is valid
parameter space a
guess

for parameters .
 paramd-ricest-imationi.IN this
problem
we want to identify the parameter
values of the distribution from

the given date set .

N-nparamd-icahmaion.IN this problem

we want to
identify pdf or CDF

directly from the data .

Examples :

Data in 2 -
D plane represented by *
.

÷÷#
A

bivariate normal density

Data comes from
Parametric estimations
Parameters to be estimated thing ?rd ,
9
 E"

.la?........Y# -

Date I D
represented by *
-

parametric estimation I normal density )

parameters to be estimated :
14,62

estimation
Ex Non
parametric
i
. -

Doth 2 - D .

:÷÷
Aqa
← *
&
*
* *
-

/
.

-
 want to learn
→

Point estimation in
parametric setup .

1) parameter estimation C already understood )

a) Definition of estimator and estimate .

brood estimator
s
) properties of an .

Methods of estimation
4)
( MOM / MLE )
confused
)
to be
Definition : Statistic
-
( Not
with statistics

A statistic is a function of data

but it is free from unknown

any
parameter .

let Xi Xn be random
EIS a
- -

:
-

,
,

sample from N Iµ or )

④ let be known
µ .

2
is statistic
I Hi a)-
a .

i= I
27 µ is unknown .

a
'
statistic
E Hi µ )
- is NIT a .

E- I
n

2
is statistic
However EX F) a
.

-
.

,
E- I

Assume M
or unknown
3)
.

{
2
( Xi -

F) Not a
is
[ = a

-
statistic
J2

Let a Xz - - -

,
X ,
du fol )
, ,

where fo tf :{to lot ④

}
objective : We
may
be interested to

estimate a function of parameter

O denoted 910 )
, by .
 Let Xi Xn be random
t± - - -
a
, ,

sample from Nlpe ,

Here D= Cµ
,
r2 ] :
parameter vector

gf 14,02 ) =

µ
C mean )

oh ( variance )
g ( 14,02 ) =

91M ,r7= I if peto

A
C coefficient 8
variance )
let Xr be random
E± Xi , ,
- -

,
Xn a

sample from GIL ,

d)

glad ) = I
×

glad ) = I
I

glad ) = A
 Definition

Estimator : Let Xi - - .

Xn be a random
, ,
-

from fol )
sample
.

statistic TCX when used

A ,
. -
.
,xn )
to estimate a parameter function 910)

is called an estimator 8 910 ) .

Node :
Tlxnyxrg - -

,
Xn ) being a

function of random variable is

itself a random variable .

EI .
Recall the example of collecting
data tossing
'

by times
'
a coin n

and then estimate the

probability
I p)
of success .
.

Let Xi . - -

,
Xn be a random sample
,

from Bernoulifp ) .

Here parameter a =p .

Xml I
910 ) =p TCA is
.
-
. =
an
,
,

estimator of p .

B
Definition ( Estimate ) .

The value of TCA

,
. - -

,
Xn ) evaluated

at the
given date points * ,
=
ni Xz -
- see
, ,

-
-
-

Xn= an is called as an

estimate .

Examples of estimators .

Let Xi Xz -
-
-
Xn be a
random
, , ,

sample from fol ) . with

Elxi ) µ for
=
it ,2 ,
- - -

in .

varia .

)= r2

Then µ
can be estimated by
i
) TCH ,
-
- -

,
xn ) =
X ,

si
) TCA ,
-
- - .

, xp = 5%-7×92
12
n
iii ) tch
,
.
- -

,Xn7= Eaixi
F- I

where
II ai = I C ai E foil )
iv ) TCM ,
.
-
.

in )= I
 E 02 can be estimated by

silk xn )
in ! ki IT
- - -
-

, =
,

1h

sick ,
- - -

,xn7=
I E. IT
Hi -

Propertiesofeslimators
Unbiased estimator : Let TCA
,
- -
-

,
xn )

be an estimator of 910 ) .
which

satisfies

ECTCX , ,
-
. -

xn ) ) =
910 ) to c- ⑦

then TCA
,
- -
.

,
Xn ) is said to be

an unbiased estimator 8 glo ) .

E Let Xs Xz -
-
.

,
Xn be a
, ,

random sample from fol .

Let µ= Elxi ) for ist 2

, ,
- -

on
ore varcxi )
 We want to estimate µ .

Let fo =
N ( µ ,
or )
D
TCA ) X : unbiased
- -
.
xn =
, , ,

ECtHr ,
- - -

,
xn ) ) =
Efx ,
) =
µ

I
2) TCH
,
- -
-

,
xn ) =

E ( TH ,
-
-
-

,
xn ) ) =
ECE )

=
Efx
xn
)

put +
µ
- - -

= MI
n

unbiased
fu
:
=

3) TCA ,
- - -

, xn ) = 3×1 : biased
estimator
ECTCX ,
,
.
- -

,
# = 3M 4Th unless
unless µ=0 Tho
Definition : Mean
Squared Error C Mst )
of
let GIO ) be a function parameters
which is estimated estimator
by an

TIX
,
-
- -

,
Xn ) .
Then the MSE of the

estimator for glo ) is defined as

MSE ( TIX ,
-
- -

,
xn ) )

CITTA
?

= E
, )-

glo
-
.
,xn - ) )
MSEITCXI ,
- -
-

,
Xn ) )

=
Var
( Tca ,
- -

.
,Xn ) ) .

?
+ ias C Tex
@ ,
.
- -

, xn ) )
 Estimation -
:

- Estimator I statistic

- unbiased ness

Methods of estimation .

e) Method of moments .

observations lddtal random sample

{ X
,

from
=
Ri
,

the
Xz = K2
,

distribution
-
- -

,
Xn

foe ,
-
-
Rn
}
}
Given

)
we want to estimate the parameters Q .

Oi
,
-
- -

,
On where D= for , . - .

,
On ) is

the vector
parameter
Ego :
Compute
1) Theoretical moments from the pdf fol ) .

moments from the date

2) Empirical
Construct k if have
3) equations you

k unknown parameters
4) Solve for k parameters .
Example :

Let Xi Xr -
- -

Xn be a random
, , ,

Gamma Cd a)
sample from ,
.

method of moments estimate

Using ,

a & X .

SI : 1) Theoretical moments .

12,4 )
£
Gamma
mean
f =

variance Gamma 12,47

f ¥
=

frogs
moments date
a)
Empirical the .

÷
sample mean = F= I Xi

sample variance =

in II. IT
*
-

MIM estimators ⇒

I =I ; ¥ =
hi Hi - IT
 Let 3

64L
Example : o O 01 2.89 4. 41,0
. .
.

, , ,

be a random sample from Gamma I d. d) .

method of moments estimate add

Using .

Sole :

I 65
Je =
r
=
21,1

§
'

f- ,
Cai - E ) = 1.713 =
4×2

a
¥ £
- .
= 1.713
Ty

1.713
⇒ ! .
65
In =

?÷}
I 1.038
=) ¥ = =

£=l
⇒I=÷,s=o
¥ =
I .

65 ⇒ 2=1.65 . X = 1.6.55×0-96
 let 1.7 -2-1,0-8
Example : o -3
,
- o .
I
, ,

)
'

be a random sample from Nyu ,

o
.

Estimate r2 method
µ
&
using 8

moments .

sin :

mean f NIM ,o2 ) =

µ
variance f wcpr.fr/--r2
2

µ
-
- I
,
oh -

In En Ki I -
I
.

-
- O .
12 IT
72=1-58
,

let ) be
1,0 , 0,1 91,1 a
Example :
, 1,1 , ,

random from (p ) Use

sample Bernoulli .

MOM to estimate p .

Sod :
ph=o .
7

Theoretical moments should exist in

Not
:

order to use MOM .

 Maximum Likelihood Estimation

let Xi Xz -
-

, Xn be a random
,
,

sample from fol .

) .
Then the
joint
density of Xi ,
- .
-

,
Xn is given by

Fx
,
. . . ,×n( e
's
- - -

an ) =
II to
Cni )

The likelihood function .

110 ) =

It ,
focni ) =

II ,
fld
Ini )

Maximum likelihood estimate 8 O

,

IME = arg Max eco )

Of ④

a
avg
max log ( Ilo ) )
ot④
Examples :

random
Let Xi ,
- - -

,
Xn
be a
sample
from N Iµ , 1) .

Find the MLE 8 µ .

elm ) = IT fudai )
I 7

where folni ) a
N
44,1 )
Ichimy
2
-

2¥
-

= e

Acri LA
-

n '
h Cri -

at
MM) IT
¥
-

=
e
,
it
'
k FE Cni -
yup
e-
=

¥,n ,

CHIMI
'

↳ film ) ) =
-

F- log I -

I E. -
These maximizing by ( elm ) is

to minimizing sum 8
equivalent
as shown below
squares
.

avg max log ( algal )

A

= argyuin
E CAsup
-

'

cost function =
II. I Xi -

in )

did In Hi -
at = -
2
Eich .
-

su )
-
- o

⇒ Elli
I Is
=
nfu

/i=Ef
 Let Xi Xz - - -

Xn be a random
, , ,

from or )
sample Nyu ,

unknown
Assumption :
µ
:

f : known

Xi )
Recall :
~ Nco ,
i

for 521,2 ,
-
-
-

,
n .

and Xi N
if
N ( nil ,
not )

Exi
I =
n
NCM , In )

in ,

#
µ
 ¥÷t÷
E

µ§#fIt3rrn
'

Ss .
d .
rule 8 normal density

PC µ 3£
I E
µ -13£ ) Er o 997
-

-
s

÷÷:÷÷÷÷ .

to the
right
8 Za .

Let 2N
N Coa )

P ( 27 ,
Za ) =L
 In NIM ,
Mn )

I fi to )
-

2 =
N N ,
I
⇒

pdf 8 Nco , I )

*#÷¥#¥
Prob -42 Prob =
212

If is clear that

p( 27 Zi -

an ) = I -212

⇒ PLZ s z ) .
as lgdsyaemmtotss
8 Non
, .az
)
⇒ P ( 2 E -

Zaw ) =
42

Thus
,
p f Zag , E 2 E 2h12 ) = I -
d
 E II E 2-42 ) I a
Pf 2-42
-

=
-

own

⇒
I +242 Ern
)
E
PCI -2
I
Zak E fl
-

Ern
- -

Take 2=0 -

f- o 05
.

I 65
2-42
-

It
P
( I -
1.65
Irn E
fu
's
1.65%1=0.9

100cL d) % -
confidence interval on
µ
when oh is known for NLM , f)
L I 2-42 Ern
-

b I -12-42
=
Ern
Example : Let 0.3 -1.8 I .
9 0.01 - 0 .

3
, , , ,
,

o .

I be a random sample
confidence
from N (µ ,
I
) .
Find 90%

interval on
µ .

Sod : I -
2 = o . 9

2=0 .
I

0.05
42 =

Zak = I .
65

Here 5=1 known

,

h= 6

I = o .
035

L =
I -

Zak I
rn

= o -

035 -
1.65×1 =
- 0.64

Tr
It
U =
24 , Ira = 0-71

C I
f- o 64 U 71 )
.
.
-
.

,
E for the compute 99%
sample ,

CT on
fu
. .

Solna i
-

I -
D= o .
99

2=0 .
01

£ =
o .
005

2-42=2-58 ¢P(2EZ%J=
0.995

L =
I -

242 I
rn

= 0.035 -
2-58×1 =
-1.04
if

1.09
0.035+2-58×12
=
U =

CI =
f- 404
,
1.09 ]
 the
Length 8 CI

L =
I -
74 , Tog

U =
It Zak Trn

U -
L = 2 2-42 I
rn
-

Confidence Interval on r2 for NCA ,r2 ) .

Let Xi Xz -
- -

,
Xin
be a
, ,

random sample from NCM ,

Assumption ,
µ
& oh unknown .

Xi
fr
N 10,1 ) I -_
1,2 in
-
-

~ ,

2
(Xi -

a)
ref 's , 's ) -
-
I ,
=
for 5=1,2 , in
-
-

i-rz-nrln.tn/=xn
E )2gu
Hi -
Consider
?

fi In
2
i # -
M) =
[ ki E)
-

tix -

gu ) )
,
,

'

Ki E) 't CI µ ,
-
-

=
.

+
I
t.EE al
'

E.
'

ME
E. Ki Hi Is
- -

E. Hi - sup =
§ ,
Hi -
Fit n CI -

sup

=E* try
2

Ifi
m2 CI µ )
-

J2
I
I
Edna I
In ,
 I.
2
Thus Hi E)
In
-

~ ,

pdf 8 In -
i

-
I
o
n -
I
,
I -
212 In -1,42

Percent points
PL In .

,
7 Int %) ,
=
42

P ( In -

,
7 In .
i
,
i -

Ha )
=
I -
H2

⇒ PC In E In .in -
ah
) =
42
.
,
Thus ,

Pl In ,** ,
←
Int E In ,% ) =
I -
d

FCK , , .

%
'
Eiht
's In .mn/--hn

r2sE7e×i-× 1=1-2
piEcxi 1%-1,42
⇐

2%-1,1-212

Example : Let 0.3 I -9 I

8,0 ol -0.3
.
- .

, ,
, ,

O .
I be a random sample
from NIM ,r4 .
Construct 907 confidence

interval for r ?

Sod : I -
2=0.9 Here n
-

-
6

A I 1=5
O
⇒ n
-
.
-
-

2/2=0.05
 ¥¥
Is
I

$5,0 .
$5,0 .gs .
05

Plots 3 Is , .os ) -

-
o .

OS
T
11.07

$5,095 = 1. IS

?kiF)
2

↳ 0.635
-

, =
-

II. OF

u -iE = am
I -
15

-
 Confidence interval on
µ
when

r2 is unknown for Nyu ,r2 ) population .

Let Xi Xz .
- -

Xn be a random
, , ,

Sample from N ( µ ,
62 ) .

'
Assumption Both do are unknown
µ
: .

Objective : Confidence interval on

µ .

E.
2

Ki E)
'
Recall :
s =
-

m -
I

is an estimator 8 r ?

Titus
Cs 'm
"siId
estimator fry ,

in
In the statistic II the

④rn

previous case C oh known ) replace T by

s .
 :*
"

" " "" an ft

statistic :

II ta =N
Tse
. . I

IFT independence g

:*
=
" " " " " "

÷
.

I
"
, "

*¥i¥÷
42 I

tnt
-

,Hz .
The statement
probability :

Pf tny.az - E
SEI E the ,% ) =/ -
d

Ittnt .us/rn)-- It
P
( E -

ta .

san % SHE ,

100ft 2) To -

on
Iu
when T is unknown

[ = I -

think Son

o =
It tn yah -
Shh

[ HUD
EI Let 3 -1.8 I 9 O
01 o 3 O I
-

o
.
. . - -

, , , , ,

be a sample from N 44,62 ) .

construct go % C .

T .
on
µ .

I =

fixings
'

II. IT s
Hi
-

-
=

n
-
I

O .
118
=
 tnt ,
212 '

Here n=6 n - 1=5

,

I -
D= 0.9

4=0 .
I
,
⇒ 42 = 0-05

£5,0 .
05
= 2-01

06

11815¥
Oi O
2.01 x -
.
=
L = o .

035 -

U = o -

035
2-01×0.118/58
f- =
0.131

-
 Testing of hypothesis
Ho :
Null hypothesis .

Hs Alternative hypothesis
.

Let Xi Xz -
-
.

, Xn be a random
, ,

sample from Nhu ,

where
'
is unknown & is known
µ of
.

Ho µ to
-

}
-

two .
sided test .

Hi :
put No

spy }
Ho left failed
:
µ= 140 .

one sided
.
test .

Hi :
fu

No
}
Ho
fu right tailed
-
:
-

sided test
Hin .

µ > µ , me - -
 I -

µ N lost )
N

⇒
If its true µ
is luo
=

I No N 10,1 )
-

TE

±¥÷
:
" "
 of
gu )
Test on mean ( normal population
when 82 is known -

Ho :
µ= No
It ,
:
Mt No
Under the assumption that Ho is

true ,

II

Arn
so
~ Nfo ,
, )

Then at the level of significance L

,

check if the value of the test

statistic II is greater than

Trn
less than 2212
Zqz
-

Or .

Then Reject the null hypothesis .

If E F- NO then
-
Zaz E 242 ,
we

accept Fn
fail to
reject Ho .

-
Example
:

Let 03 -
1.8 I -9 -
O -
3 O -
01 o .
I be
, , , , ,

a random sample from NIM ,

I ) .

Test whether the population mean

is I at Oil level of significance .

Sod :
Ho :
fu
- I

It ,
:
petal
The level of Significance 2=0.1

Under Ho the the value of the

,

test statistic II. is

ran

0.035 - I
-
I - 2.4

yrs

check if this value lies in f- I. 65,165J

since -2.4C - libs ⇒ Ho is
rejected .