DISCRETE RANDOM VARIABLES

A Random Variable associates a value to every possible

outcome

Mathematically function from Sample space s to

a
the real numbers a random variable can take
Discrete or continuous Values

We have several random variables defined on the

can
same sample space A function of one or several
random variables is also a random variable

Probability Mass Function Discrete Random Variable X

of
It is the Probability Law or Probability Distributionof
X then X x is an
event If we
fix some x

PED P WE Rst XCW xD

pyo
p a 20 or
Ipx 64 1
n't
X
s a plab7 Ya
s
plot
i a pld t
d
3

Probability
I
Calculating PMF
For two rolls of a tetrahedral die ie H faced die
9
5 6 7 8
3 4 5 6 7
4 SecondRoll
of Die 2
3 4 56
I 2 3 4 5
Rod Die
X First of
Let every possible outcome have a probability of 1 16
Z X Y

to find plz
collect all possible outcomes which 2 3
add their probabilities for

pay

o
g illiii
 Bernoulli's Random Variable Parameter
with p Eco
I up p
x
o
up t t
It models a trial that results in either a or
H or T etc
Indicator Random Variable of an event A
L A
Ia iff occurs

Discrete Random Variable params a 2b

Uniform
Parameters integers a b as b

Experiment Pick one

of a at 1 b at random
all equally likely

Sample Space a atl b b att possible values

Random Variable X X w W

Model
of complete ignorance
a a
special case a b
bin pyx

It I

a att b K a 2
 Binomial Random Variable

Param positive integer n

PE 0,1
tosses a coin with
Experiment n independent
PCHeads of
p
Random Variable X number of Heads observed

Model
of Number
of successes in a
given number of
independent trials

n
p
I P 3
HHT
p ni
IP P HTH
p go 2

p TH H 1
t t b
pp
THT go
Ip
TTH 700 O
TTT

Px Ck 1 place pink K O I
for
n
 Geometric Random Variable

Param O
p pet
Experiment Infinitely many independent tosses of a coin
P Heads
p
Sample space Set of infinite sequences of H T

Random Variables X no
of
tosses until the first Heads
Model
of waiting times no
of trials until a success

Px k P no heads ever

pxlb p lb

I a s n s d e to

8fection Mean a Random Variable

of
gain at
Motivation 1000 times Random
Play a
game
each described
play is
by

I 2 wp 5110
 I
Interpretation Average in large number of independent
repetitions of the experiment

Definition EG I xp x

Caution
If we have infinite sum it needs to be
well defined

We assume
I 14pyle a

Expectation of a Bernoulli's Random Variables

to up p
mo p Ip
X is the indicator event A
If of an

Expectation of a Uniform Random Variable

Efx Expxco

Elementary Properties of Ecpectation

Y X20 then EXT 20

as x sb then EEXTeb a constant
Ifthen ECC C
as
if e is
The expected value rule calculating ELg x
for
Let X be a v.v and let y g x
ELY
Averaging our
g Jyp ly
Averaging our x ELY E Eg X
Eg 2 px x

In general Ecg X EX
g
Linearity of Expectation E Lax b a ECxJtb

n
g 5
or 5
0.3 so 4
p
prob
3 3.3
25 o 2

Variance A measure
of the spread of a PMF
Random Variable X with mean M EXT
Distance mean X te
from

Variance var X E CX 1413

Standard Deviation Ox ay
Properties Variance
of
var lax b a'var x
Notation
ME EX
Let Y Xtb where Y ax

var X EXT CENT

of Bernoulli's
Variance

x
to up p
w.p I p
var x EG EET px x

var x E X EXT
Conditional PMF and expectation

Condition on an event A use conditional probability

px x P xx pya CA PIX HA
Ipx x I E pXIA D I

EEx EAX x ELNA Expxia x

EcgCx Eg a PXCD Ecg X AJ Egenpxla x
 Total Expectation Theorem
AnB
A PCB PCA PCBA t Plan PCBAn
a
At

AT AzOB

for fox is
Similarly x

Px a PCA
px A x t Plan PX An Ca

EX PRADECXIA It PLANE EX An

Conditioning A Geometric Random Variable

X no
of independent coin tosses until first head PCH
p
Px k I f he 1,2
p
pick Memorylessness No of
remaining
coin tosses conditioned on Tails in
the first toss is geometric
with parameter p
I m

Conditioned on X 1 X 1 is geometric with parameter

p
The mean
of geometric
EX Ekpx Ck ÉkCI pimp
EXT
Ip
Multiple Random Variable and joint PMF's
x px
Y py
Joint PMF x Px x
24 y
px y y
I
Eye PX Y x
y
Y x
foxed Ipx y
pyly Efx 4 x y
The mean
of Binomial
X binomial with parameters n
p
no
of successes in n independent trials
where Xi I ith trial
ifotherwise success
Xi O
X X Xat Xm
 EX É k
I p Cy pink
or in simple words EXT np

Conditional PME
PYA HA P X KIA foxy PIX MY y
Hy
such that pyly so
fay Hy [Link]
defined
for y

Ep xx Hy 1

Conditional PME involving more than two R V

pay z k ly
z on
fx y1z x
yz
we can simplify with regards to Multiplication Rule

PCAABAC PCA P BIA P CIANB

Px4,2 x
yz fox x
pyx Cylon fax 212y
Conditional Expectation

EX Exp xGc ECHAT Expxia x EE Y yFExpxiyG4y

The expected Value rule

ECG x E gG4pxGD [Link]

Ecg x y y EgGopxlyGuy
Total Probability and Expectation Theorem

Ai An partition of a

Px 2 PCA
pya a t Plan pxlAnCx
EX EX Y y
Jpy y
it is also valid when Y is a discrete RV that ranges
over an infinite net as long as ECX1 a

Independence

of two events PLANB PLA PCB

of a R V and an event PIX x A P X D PLA Fx

of two R V's PCx x2Y y Px x PLY y it x

y
f x
[Link] pxcopy y y

X Y Z are independent
if
2 A n
fox 4,2 x
yz px x pyly pz yz
Independence 1 Expectation

In general E Eg Xy
g CECI ELY
Exceptions Elaxt b aE x tb

E XtY ZJ EEP ELY EEZ

If X Y are independent ECXYJ ELSECY

also
g x and h ly are independent

Ecg Ahly ECG x ECKCYD

Independence Variances

Always true var ax a'var x var xta var x

In general var xty var X var y

if X Y are independent var City varix var Y

Variance
of the Binomial
X binomial with parameters n f
number of success in n independent trials

Xi I ith trial
Xi 0
if otherwise is a success

X X Xat Xm
 Continuous Random Variables

A R.V is continuous
if it can be described
by
a PDF

pain

Plas XE b

Plas X Ib [Link]
bxG pxlnlzoEpxCd
1 fyi20 ffkldx l
Plaexsb dx
fx sc

Probability Density Function PDF

Places at 8 Eff Ca 8 Plas Xsb f fxcudx
P x a O
Jfk 20

1 feudal
Expectation Mean of a continuous random variable

Interpretation Average in large no of independent

repetitions of the experiment

EX Expx x EEN
f xfx Goda
assume
[Link]
Properties Expectations
of
X 20 then EX 20
J
X sb then as EXT Eb
J as

Efected Value Rule

Ecg x
EgGDfoxed I get fx Code
E x'J J X
f Goda

Linearity
ElaXtb a EETtb
 Variance and its Properties

Definition var x ECCX 111 T where

H ECX
value rule
Calculation
using the expected

EEg x
Igloofx Code
var x
12 x a f Golda gac x
pet
Standard Deviation ox Tax
var lax tb a'var x

var X EXT CECH

Continuous Uniform Random Variable params a b

find a paid n

bat
I I a att

EXT ALL
b

E x
f xfxGodie
du
fab no Ia att
EXT Joe b
Fa
da
ya g
 EXT EXT
var X

Standard Deviation
BEI
big
t

Exponential Random Variable params X'so

fixed fx Go

E
Ci p p

I tr

te azo
fan xc o
O

E x
Tx Xé doc
I
E X
Jail e dx
Ig
var x E X CE XJ

PC x2 do
a
fixe
ea a s t
feudal at
t C e
ng
ta
e [Link] e
lead
Standard Normal Gaussian Random Variable

Standard Normal N O 1

J die ran
fx x e e

EX O
var x 1

x
General Normal Gaussian Random Variable

2 Ck mat
General NCpe 0
fx x
fye
EX he war x o

A A
normal o smaller a

Linear Function normal random variable

of a

Let y ax tb X V N
fu o

ELY qutb
var Y aka
Fact Yu NCalutb att

Conditional PDF an event

given
fox x P xx N S E PCRSX ext g
P
yay SEP KEX 8
f A
EX
Pxia x XIA a
 PHEBE EBpx P XE B Goda
Ifx
x

PCXEBIA ZapXIA CD PCXEBIA Gtxlaced

Epxialic 7 ft a codicil
conditional PDF X that XEA
of given

PLEX IN SIX EA Efx YEA CK S

[Link]
PCxspcExtIItXpEy
fxixen is A

yo
FIT if Kea

Conditional Expectation event

of X given an

EX Espy Cio EXT fxfxcudx

E XIA Eapxia Gc ECXIAT xia Gada

 Expected Value Rule
E glx I glopx Gc ELglxD
fgGdfxao sEEgCxllAT
EgGopxla [Link] [Link]

Memorylessness
of the coponential PDF
Do prefer a used or a new coponential
you
bulb
light
Bulle lifetime T cofonential X
PCT D é 220
for
we are told that t t

r v X remaining lifetime T t

Now

p x KIT t E for 220

PCT
[Link] PltottEn
t
eF
e
PCEEE
 X
Xe 220
fila for
PCOSTS 8 a FCO S XS
PCETEESIT E ITS

Total Probability Expectation Theorem

A AND
PCB Pla PLBApt PLAN PCBAm
A AND
[Link] a Got tPCAmlpxIAC2
at Aonb
p yea P CAPP XENA
fx a

PCA
t

FLA 2 t

PCAR
EXAM
EX PCADEENAJ PLAMIECHAM
PLAN
ECXIAL
PLAN
E XIAO

lg Bill goes to the supermarket shortly with prole

13 at a time uniformly distributed b w 0 2
hours or with probability 2 3 later in the
day at a time uniformly distributed b w 6 0
hours from now
 PCA
as
31 fila unit 0,2

P A2 23 fxia V
unit 6,83

EX PCADECHAI PCADEGLAD

x1 t 7
3 n
43

It 116

IT o

Mixed Distributions

X on 0,27 with prob 112

ugiform
with prob 112

Y is discrete wilt prob p

z a continuous I with prob I p

FxGc PCYe Nt CI p PC D
f
pFyGot I p Ecu
ECXJ PECYTT CI P JEEZ
 X uniform on
0,27 with prob ya CA
with prob 112 Aa

Fx a [Link] PCAA FIA CD

Fxlan FXAyla

l 112 l
ppg

o I o

FxCu
Slope Yy

y
4a slope

I I
0
I 2 I

Jointly Continuous R V's and joint PDF

Two random variables can be jointly continuous

be described PDF
if they can by a
joint
foxy G g P X x 24 4720 Jx y Cosy 20

PAX y EB Ey É pay Cosy Plex41GB fxyGpd

f
CgEB
I l
EFpx y Cosy [Link]

On Joint PDFs
PAX Y EB G y dxdy
f
GYIEB
fx

PlasXeb Eyed f fx y Cx g dady

Plast sat 8 EYE ct872fx y a C 82
[Link] probability per unit area

Conditional PDF's another v.v

given

pxiycxlylPCX nly 5 [Link] ifpyly

o

fayGuy [Link]
iffylyls A PGcEXE2t8
8 where PLA 0
Zfx1ACK
Plexsutsly SY Sgt E I
[Link] fxyG4y

Definition PCXEA14 91
f fay rly dx

Comments on conditional PDF's

fay Hy friggin ffxly aly 20

Think of value of Y as fixed at some shape of

y
[Link] slice the joint

I fxiycuyldx f [Link] l
g
Multiplication Rule

[Link] fyly fay rly

fyou fyxLyla
Total Probability and Expectation Theorem

Axa Jpy y pxly rly

EEXly y Expxly Hy
EXT JpyLy E XIY y
fxcu f f y ly fay Hy dy
E CHY y Fa fay ily dx
Iffy y E X y y dy

J fyly Ifafmy rly dn dy

fix f fyly fay Ny dyda

f a friend die Efx

Independence

Px y ay px a byCy f x y
fay ay fxGa fyly t my
fx y la y fxly Caly fy y
equivalent to fay Hy f xGc t y with fyly so lax

YX Y are independent

EX YF EEXTENT
var x 4 var X van y
 glx 2h14 are also independent

Elgexthly EIGHT ECHLY

Stick Breaking Example

Y d
d le
Break a stick of length l twice
first break at X uniform in Co LJ
second break at Y uniform in O D

fay ay [Link] I
fun n

e n x
y