Mean and Variance

Mx(t) yet ; E(X) My(o)

= = = ze = T

Mx"(t) Net ; ELXT MY(o)

= = Fe T =

Var(X) EiXYfE] -E
FLtF)
= =
=

**
that PC =* LH valid
Verify is a

prf
=
= r + x
=

valid
:
p(X) is a
panf
Binomial Distribution

X =
the number
of success in in trial

parameter
= success
probability ,
i and
In

XvBin (n, 5)

in
E
a

otherwise

for X, N

let Xi BernoulliCi)
mgf where i 1
~
: =
,
2
,
3 .... n

xi

Convex function .

g(x) D
(
g(xX x)xy)
,
+ -

get
*

xg(x)+(1 *)g(x2) 8

if g(xxi + (1- *) X2)

xq(x) + 117)g(x2)
-

them
g(x) is convex

X
Y
Xz
XI+Xe
where o <
x/

Elg(x)) g(E) ,
if g(x) is convex

Jenson E(XY) >, ELX) ·

ELY)
Inequality E(k)3 EL ,

Binomial Distribution

X the number
of trials
=
success in 'n/

~ Bin (n , M) T
= success
prob.

N
Sexn a

otherwise

&
Check
validity of
> the
ps
-

=
(r+C ,
(2)(ft +QUI ... + =+"=

#GymBernoullic) where i 14: in

N
++ .. -

+ X win)

,
7
00 .

750 -

--Yu x = + %2 +For

X
=Z Y Bin (n
, 2)

Mx(t)
My (t) My
=
, (4) yi's are iid.

(f)
(My

=
(re e +
Poisson Distribution
X the number
of
=

occurances
of an event in a
given space
parameter = the
average
number
of occurances
of an event
per
unit time (per unit
of space)
(4)

pmf
Y
ex x
:
p , (x
=

[
.

xI
for M = 0
,
1
,
2
,
-..

j otherwise

The
validity of puf
Px(X) ,
o

=
c(Y)
=
eb(Hx+ + + . . .y

)
= e ex e 1. Px(X)
.
= = : is a valid puf .

MGF
of X-Poisson (i)

Mx(t = Ecext) = ext X

= X
=[1 xet Let
]
+
...

=
-X exet get ) . =

Myct) excet 1) yet

-

= .

Mx()
*4
1) xe" 1 x 7
E(x) e
= =
.
= =
.

)
Mxist) excet-= .

xet. xet + excety .

xet
C. xe: xe + *)
ECXY Mico) =
e e yet
-

=
y+ X

Var(X) E(xY = -

[E]= +x - ** =
x

ECX) and Varcx) by

using c)
(x(t)
logMx(t) logexet x(et )
=
1) =
-

=
-

kx(t) = xet ; E(x) ky() xex =

=

KY(t) Yet =
; Var(X) kcos xe" = =
=
x

Hyper geometric Distribution

A lot of Nitems has K defectives n items are selected at random

1 2 3 --- N
, , -
Is defectives N-1 non-defectives
↓ selectn
↓
defectives and n-x non defectives

X the number
of defectives sample of
=

in a

Xe Hyper (n , N K)
,

prof PLX) = 13x : Pax

Nen for X =
1
,
2
, ...
N

E(X) = Koki
Nom

-
K!

n -
1

IX- We N - ka

=
(N-1) !
D! (N -2) ! (m- ! N -
10n-

=CXN The
prf of X-1
,
X-1-hyper (n+, N +, (1)

=
Ex i =
 Geometric Distribution
·

Version I

* The number
of failures before the first success

FFF
... S
parameter
F =
success .
prob
First success

X-Geo(T)

P,)
=
(H iv for x= 0
,
1 2,
, ---,
0

P otherwise

·
To
prove
the
validity of P*)

Px(x) , o

=+(H)(H) +

==
: .

P(4) is a valid
prof .

=> Fx(x) =
P(x = X)

=
x+ (r 7)(x)+ (7) + --
- -

+)H)7

FAA
mgf Mx(t) =
Petty (examples)

Version 2
of Geometric Distribution

X the trial number

=

of the first success

H forX
I
Ex(X) ,
= ·

of
Fx(x) p(X(X) px() = =
+ Px(z) + ---

Px(x)
*
z+
=

(17)7 + (1) z + - -
-
-

+ (1 -

2) 7

7 X terms

NEH
Memoryless Property ↑

H Success
=

TTTT
-

P(more than 6 times to

winl more than 4 times to
win)
=
P (more than 2 times to
winl
=
P(4 + 2)4) =
P(2)
P(x4 2)x + > 4) P(x)2)
=

- P(x)x y)xxx) P(x+ =

>
y)

method"
"
000 Sometimes Mxio) and Mxio) doesn't work to find ECX) and ELXY
,

In such cases we use

,

series
-

power expansion
-

ECX) = the coefficient of

ELX) =
" 1 2
!

Negative Binomial Distribution

"X the = trial number of

th
success

Trials : 1
,
2
,
3, . . .
X

u
2-1 successes oth success
in X1 trials

P xCr I =

for K =
W
, ,
+z ,
-
-.

X-Neg ·
Bin (r
, x)
& What is the distribution
of Xif X-Neg Bin(17) ? .

Wecansa it XGl th
success.

Trial
-
Total =xtr-1) * failures ,
2-1 successes oth success
 P(X) =

x+ - 1 Cr- 1 (x) ! (ty) * x

Common Continuous Distributions

Uniform Distribution

k-- X-V(a , b)

a b Derive
fx(x)
b

S(xx)dx =
- Skax ) [kX] =
1 =

( kb-a) = 1

= k a
=

Fy for axdo
- :
a

otherwise

) Fx(x) PLXIX)
=

= adt

Se
Fx(x) =
=o

mgf