STAT 1202

Statistics
-

Introductory -
LECTURE 1 : INTRODUCTION TO STATISTICS

SUMMATION NOTATIONS POPULATION & SAMPLES

→ index ends at 10 -

mr/ eaf.name/lPfnrea.maemt;Yeot#-
10
*

}
q i
= I t 2 + 3 . . - t 9 t 10
N =
Population mean

parameters of the
↳
⇐I i 's
population variance
'

summing population
02 =

↳ index starts from 1

( O
-
-
standard deviation )

iii. iz = o
.
+ x. it .
. . . x. so
.
Tiffani:

MARGIN OF ERROR

Difference d sample
→ between population mean mean

FORMULA :
2¥ / z,Pxln
↳ * of selected ( p) to find
probability
being
throb ability varnishing
of
Thee
need the

MOE margin
error

: non - -
cannot be calculated )

equal &
→
every member of the population has an known
PROBABILITY SAMPLING
probability of being selected that is to
°
Random selected
randomly another member
equal chance of
being selected independent
-

, on

being selected
°
System e tic -
first member selected from the sampling frame at random ,
using a random number

°
stratified ↳
I chance of being selected ( Eg : to choose 100 people out of 500 ,
randomly choose a

÷ starting point & select

every 5th person )

NON -

PROBABILITY SAMPLING ( not ever body has a chance )

°
Quota ( to compare 2 diff
groups
in the populate eg :
males vs
females )
↳ sample will be proportional to actual population

°
Convenience

↳
choosing samples based on ease
of accessibility leg : First 100
people who walk into school )

°
Purposive / Judgemental 1 Subjective

↳ specific goal
choosing samples e a

°
Self selection
-

↳ participants choose to be part of the sample themselves

Snowball
sampling
°

↳
finding specific hard to reach samples leg :
drug addicts)
 ERRORS
DISCRETE

DATA
/
-
NON -

SAMPLING
-
NUMERICAL ( countable,

1
Imperfect research
design

/
-
CONTINUOUS

#annan}¥'d :{vetoed
unearned '
Mistake in execution
1 1
-

related to ttlargersamplesize
Any
,
-
other reason not
SAMPLING difficult to quantify CATEGORICAL
sample selection

\
ERROR

/
( no order ( ranking )

Margin of Error qualitative ( non numerical )

-
-
-

°
Systematic Error

imperfect aspect of the research

-

/ any factor not related to the

sample selection
HOW NOMINAL DATA -

categorical qualitative leg

.

gender blood type

REDUCE
,

TO
- , ,

postal code)
larger sample
-

size

Estimate population 0
Random Errors
-

ORDINAL DATA ordered ( has ranking

confidence categories lorderieg
-

parameters
:
a

intervals managerial positions in

a
company )

INTERVAL DATA -
differences between measurements NO

{
,

200cal 2604
GRAPHICAL PRESENTATION leg
TRUE ZERO : 150C - 210C -

Quantitative ,

continuous
OF DATA

fifteen.gg?algwe,measnrements.Truezero
RATIO DATA -

* TRUE ZERO
of property being measured
: absence the

CATEGORICAL NUMERICAL ↳ OC : no true zero -

ooo Ino heat

°
line chart ( show aralne ↳
frequency changes in
distribution ✓ true zero
°
Okelvin : -
there is no heat at 01
Time overtime )
^

#
6

ocrosstabie
Ill

l-rari.net.in#re..I7noeh!a7Iia7n7
# o class intervals
't
,÷:th
no -
of desired intervals
' interval '

-
o
frequency distribution →
intervals are
mutually exclusive

mins
intervals
ospaoingbiw )
* → max 's
Histogram (
. .
"

ORANGE 500 600 750 the bars

-

/':/¥TI#
-

Age Frequency

Htt
-

obarcna.it soo .

I
200

Hii
-
-

"

✓
137.42¥
31-36 11
>
-

30 35 40 45 50
A B C

stem & leaf display

o - -

Pie chart
°

many class intervals ( narrow)

→

:/
stem leaf ( 52,55 , 63,69 . 82,83 -
jagged distribution with
gaps from empty classes
( tens ) ( ones)

indication how
poor of frequency
-

varies
-
-
across

:
connie:*
:/ 7

8 2
! Imai : engineer → II:
-
intervals wide,
compress variation too much
distribution
.
yield blocky

- obscure impf patterns of variation

°
scatter plot ( observations taken from 2

numerical variables)
To DATA PRESENTATION ERRORS To

/
?
widths must
10

histogram interval
* be
¥ equal
- - - - - - - - - - - - -

x
! *
compressing ldiitortingverticyaxiv.no
point
-

;
Zero
( makes data diff
appear vastly
.
×

actuality
5 I
when in
there little diff )
- - - -

is
, ,
.

ulideopg 19
I

↳ see
/

:#
80
* do
30

provide basis
f- relation b/w 2
groups
< GRAPH PLOTTING ON EXCEL >

°
BAR & PIE CHARTS

i. select data

toggle)
2 .
insert ( top left

3. Recce .
charts → Choo ve the chart
you want

chart element
4 Add → axis titles

↳ x - axis : Im

↳ y axis - : Im

↳ chart title : Im

↳ correct data & presentation : 3M

5 . Add chart element → data label

TO HIGHLIGHT THE DATA OF ONE OF THE BARS :

i . Double click on the bar

you want to highlight
ii. Format data point → Fill -7 solid fill →
change
colour to make if itand out

÷÷÷÷÷÷÷i"
STACK PIE & PIVOT

/
°
.

TO PRESENT IN A CHART :

iii.sets:*::::
table
iii.
" "" "
i. ai . any data ."

2 . Insert → pivot table

"

"" "

T.ie::: as :
' "
ii's .
iiom
" e.
investment.
drag
.
→
g. '

VALUES
'

↳ variations for the columns

leg : Investor A. Investor B . Investor c) →
drag to

To change type of data presented ( arm 1 average I count Imax 1 min

etc )
.

right click field

settings you want
' '
change
'
→ → → WM to what

CHANGING B IN WIDTH

'

to show Format Data series

'
→ click on
graph

→ 3rd toggle ( Ms ) → Bins → Bin width

LECTURE 2 :
DESCRIPTIVE STATISTICS

MEAN MED BAN MODE

N =
IT 27=1 Xi
-

Ascending order -

most
frequent
Median position
-

Odd no . of values : median =

middle number

Even no .
of values : median =
average of the 2 middle numbers

mean

median
mode

Positive skew 1
it symmetrical
.

Right skewed
-
Distribution

|m
mode
median

mean ,
median

-
skew I
Negative
left skewed

QUARTILES
-

LOCATIONS
00 , 125%1 :
IT ( htt )
°
Q2 ( 500/01
}
( htt ) → median

° 03 (
75%1
!
4- ( htt )
ICR :
03 -

Qi → middle 50%

8 0.9 I O I 2 I 3 1.5 I 6 2 O
Eg
: O .
, ,
- , .
,
-
, . -
, -

Qi :
LT (8-11)=24 → 0.9 t IT (1.0-0.9) = 0.925

1.0
0.9 #

Position 2
'
14 of the Position 3
way
 BOX E WHISKERS
( Median )
Q1 Q2 03

|%/
Minimum Maximum
-
( smallest value 't outlier) ( largest value t outlier)
go % values ( ( ( R)

values > 1.5 box

lengths from Q1 103 :
OUTLIER

values > 3 box lengths from Qi 103 : EXTREME OUTLIER

median

\
①
variability
/

VARIATION

↳ population variation ( 04

↳ sample variation (
-
s I

/
62 =
Ty 8 ( sci -
N )
2
S
-
=
NITE ( sci -

E)
2

# net, F.) IZ
=

8 ( xiz , (
- =
-

N sexist -

STANDARD DEVIATION

↳ population Holder .
( O)

↳
sample Holder . CS )

• =
IT
/ s
-

-
1st
 COMPARING VARIANCES : COEFFICIENT OF VARIANCE ( CV )

describes the dispersion of variable without

a
depending the variable 's measurement unit
-

on

( relative variation )

in percentage

- Used to compare datasets measured in different units

Population CV :
f- / sample cv :

MEASURE OF GROUP DATA

:÷t÷
÷

/
POPULATION SAMPLE

¥1
" fi
N Sfi * f: frequency
=
TOTAL COUNTS

" n' """ " " "" """

e ximi ,
mis
.

mean,

62 =
# Effi ( mi 2
] S2 # Effi ( mi
VARIATION -
N ) = -
E) 2
]
'
Efimi
-

Efim 't
( II )
= =
'
N
-
I'
NT
-

µ-
LECTURE 3 :
PROBABILITY

PROBABILITY :

Measures the chance that an event will occur

A B
A

①①
B

collectively exhaustive ( events

( "" "" '

s"
"" " " " " ""

Im
""

a'
'

CLASSICAL METHOD
(
intersection ( AAB )

YIIYI.FI ÷ } (

|m÷
-

All outcomes are

equally likely to happen Union ( AUB )

°
no .
of outcomes that satisfy A . P ( AUB) =P ( Alt PCB ) -
PLAN B )
PCA ) = -

total no -
f- outcomes in sample space

A B
THE ADDITIVE RULE

!!!!
""
"

" "" " " " " " " "" "

°
PLAT -

-
I -
PCA)
(
complement ( B )
CONDITIONAL PROBABILITY

PLAN B )
°
PCAIB ) : i PCBI A) =

PLA )
°
Pl A/B) =
; PCB , A) =
B
PCB ) PLA)

-
RELATIVE FREQUENCY METHOD

classical method cannot be used )

when
leg when n approaches as
-
:

STATISTICALLY INDEPENDENT EVENTS :

One event occurring does not make the occurance of the other event
anymore Hess probable

⇐
TESTS OF INDEPENDENCE

PCA AB )
°
PIAIB ) =
-
= PCA ) * PCB ) > O

PCB )

Plan B )
=
PCB ) * PCA ) > O
PCB I A)
ta
° =

VS
Mutually exclusive :
o
PCAAB ) =
PLA ) P ( B ) -

PLA AB ) -
-
O

* If two events are

mutually
CANNOT be
exclusivethey ,

independent
WEEK 4 :
DISCRETE PROBABILITY DECISIONS

DISCRETE RANDOM VARIABLE is a variable in which the vet of possible values

consists of isolated points on the number line .

Let X be a discrete random variable and x be one

of its possible values :

Probability of X
taking
a
specific value x is P (x )
-

on -
x .

PROBABILITY DISTRIBUTION FUNCTION is a

representation of the probabilities for all the possible outcomes .

Eg
:

× p ( x
-
- x )

2
¥ = 0.25

DISCRETE RANDOM VARIABLE

(
PROPERTIES OF DISCRETE PROBABILITY '
var ( x ) = 8 (x - E Cx ) ) Pla )

IE
"
°
Same Pint YI
°
MEAN : E (X) -
-

N
-

-
E Ix Pla ) ]

DISCRETE DATA :

/ Ell X ) Ty / Ty E
2 '
E IX ) PIX )
-

N)
- '
°
VARIANCE : P (x ) -
N -

62 = E. ( sci -
N)
2
( sci 21 -

°
CUMULATIVE PROBABILITY : F- ( x ) = C ( x =
.
x ) = E x ex Plk ) = PIX EX )

DISCRETE PROBABILITY DISTRIBUTION

A. BERNOUO.at DOG 'T Rb BUTTON 21 . BHNOMOAKDOSTRB BUTTON

-

Events we 2 Outcomes ( 41N ) -

counting success I failure of 2 -

state outcomes

Mean : E ( x ) =p fixed '

- '

n
-

Variance : var (x) =

Pll -
p)
-
observations are independent
D
X -
B ( hi P )

FORMULAS
31 PASSOW DOG 'T Rb BUTTON
.
o
c : :

x ! (n -
x )!
used to determine the probability of random variable
-
a

which characterises the of 1 successes of sequences are mutually exclusive

-

no , occurrences

certain event in interval

a a
given continuous
( f- occur at the same time )
( of accidents from b
Eg no on a
highway 8PM )
-
: .

on ! = h (n -

1) In -
2) . . .
x 2X I
^
P (x) (e ) for i
PIX ) O for all other
-
°
= x -
-
-
-
-

values of X

* where it is the mean rate ( occurrence of event per unit time )

° MEAN : E (x ) =
A

°
VARIANCE : Var ( x ) -
-
X
LECTURE 5 : CONTINUOUS PROBABILITY

continuous Random Variable

X
probability density function , fix ) function of x when x
-

is a
- -
.

area under distribution ( !)

probability curve
integration
-
=

( X 42 E 2 P ( 27=0 )
-
-

value X
at specific
=
-
P a
-

-
O -

↳ * prevent how ga is
phrased

Properties of probability denvity function

① f- ( x ) so for all values of x

fall floc ) dk = 1 ( P ( all f- ( x ) =

1)

② Mean : E ( x) = faux K '

f CX ) disc

③ Variance
faux [ El XII ) fall
'
var ( X) ( ( El x ) 32
-
x f (x) dx fix ) dx
=
:
x
-
=
x
- -

④ cumulative probability :

""
n At a
point height of
. the curve : fix ) / PDF
( axis )
y
-

't
::Ii
"

!
: -
Discrete :
PMF =

probability

amiiacieiiitinemweea.int '

" " " between " '

: :: :c; : :
a '
:

a b

Uniform distribution :
equal probabilities for all equal -
width intervals within the range of the random var .

f
f- (x ) = for a Ex Eb

for all other values of x

Total probability under

graph
-
-
I
- - -
-

:::c: ÷ :i÷÷
n

-
X min -

- a Xmax -
-
b
Exponential Distribution : distribution for events that
randomly happens
leg : customer
walking into
a
shop ) .

& time between 2 consecutive transactions at the ATM )

leg
occurrences :

f- ( x ) =

X
MEAN : ECH -

-
E

VARIANCE :
Varin
L
-

'
' ' ' bi

e'
f.
a - Rb
-

.
..

A b

Normal distribution :
random variable hav an infinite theoretical
range

HEY
-

OHH
-

e
-

✓ 2162

°
Mean :
ECXI
-
-
N X -
N ( N 02 ) ,

°
Variance : Var ( x) =
E (X -
N 12=62

f- ( K )

.
N

mean
Median
Mode

Standardization :
standard normal distribution 2. where ← Nl Oil )

"
°
X → 2 : 2
-
-

IN
PC XL 8. 6) =P ( 2<8 .bg?o8iO ) PIX > 8. 6) = I -
Pcxc 8.6 )

I 0.5478
=

=P ( 210.12 )
-

→ TABLE 4
=
O - 4522
= 0.5478

Probability of Normal Distribution

q
IN
"

PILL 1=0.2 → 20% ( Table 5) Table 5 :

XZXCP )
N -
- 8

IN
"
6=5 2
=
= -
O . 8416

"
? :O : Iai: ! :L: III. I. means : adder .
-
-
sareieootnanataia .

MT x

X? 8 2

2?

✓¥
Empirical Rule

I .
NIG contains 68.26% of data

g. nt-goontain.gs?:%74::aa
Nt % of 6 contains / I ×
-
36 -
26 -

G N O 2636

Z
3 I 2 3
-
-
2 -

I 0
LECTURE 6 :
SAMPLING DISTRIBUTION

Population Sample Sampling frame error Random sampling error

✓S certain sample listed error due to sample detected :

elements not
-
-

( N 6 ) ( I S ) in the difference b/w Populations

or
accurately represented
,
,

sample result ( aka margin of

sampling frame leg :
database not error )
updated )
using sample meant
eat population mean

I -

m.O.es Ns Itm .ae

N
need the sampling frame to calculate
probability of
-

m.o.esns-xtm.ae Central limit Theorem

awinnoobnatpi.mn?o?YYalnantioYtI7aibceuiYa9teYYYp:Inia-
-

→
ion
I
zernsn -×+zEn
-

ciitf population is normally distributed

theoampling distribution of
,
-

cannot be means
estimated leg convenience sample )
.
mean :

r
Willaloobe normally distributed .

Cii ) tf population is not

normally distributed sampling
the citrates that the
I -

m.ae I + more
,

distribution 9- meant will tend

towards a
normal distribution ausample
vile increases ( hbo )

& "
'

Population mean :
N = population variance : o
n

"

8nF On %
'

sample mean :
Nz : =
µ sample variance :
GI
'
= =
{ = → STANDARD ERROR S

sampling distributions of proportions

NOT
normally distributed
-

:
only YIN

Fhwa binomial distribution but be

approximated to
I>
POPULATION can
-

SAMPLE . a

BY CLT
normal divtribntionifnpll -

p)
> 5 .

µ
,

II

µ
"

YI
"
" Tt "'
?,xi
In
=

lfntso.by CLIF
"

Nipp - op
-
I
-

. .
, , ,

62=fiE,Cxi
'

m2 ↳ = =
l l
N n
-

'
i
POPULATION A- MME
>
x ! I -

XI N ×
I, I
µ N =P
Np N p
-
-
-
-

l l
-
✓
l
G =P ( t P)
Pll
-

I N p)
g§2=
-
-

[ = I

IN
" N n
0
Z=
g- ( En )
-

al'll Pl
-

= I -

Gp

A
Pll p )
-_ -

A
"

it:
l
Ii: l
a.
Ii:
l
l
l
l

> l

N' O z
2 ?
-

2 ? N O
-

2=0
2=0

&
Point Interval estimates
-

confidence
Interval
I
p
tower 85% 90% 95% Upper
point
confidence estimate me confidence
( if Floris
Limit exactly at the confidence limit
poih-feotimafe.it ivan
Levey
unbiased estimate ) Ccl

Iisaneutimafeforlv
•

ouzisanettimateforo
'

il known population variance . o

-

( assume normal distribution )

2=1 -

Yoo
a :p
LELE )
→ .

II.
-

marginal error

in unknown population variance ,

o -
lavonne normal distribution )

l l
→
wet score distribution →
need to find degree of freedom
-

( v1 l

:µ=I±tvF( %)
-

2
→ Cl -
t
e
-

✓ =n -

I t
t

LEINE )
- F
:-( ng )
I -12
-

margin terror at
-

Yoo - -

lower limit
upper limit

Interpretation of
f- ( Ernle Ys :-( En )
the
Cl : Weare c% confident that population mean is
between I -
2 I -12 .

Confidence Interval for proportion POPULATION MEAN POPULATION PROPORTION

"' "
KEI ( g: ) CLE ) 't )
-

a- n
;
-

in fi
Cl :p
f±2E( )
e
-
-

sample vizeivlargeli.ee :npu population normally distributed

-

is
p) g)
byut
→
* ALWAYS
>

!
-

ROUNDUP
,

ol
ftp.notgivenihqneution.uoeworuf cave
possible scenario take P' to be 0.5 !
-
-
'
- - :

Reducing Margin f- Error ( LE ( Ent ) ( for population proportion )

:

+ to :
depends on
population
Th restricted due to
sample financial
X :
vice is reasons

✓ tell
LECTURE 9 : SAMPLING DISTRIBUTIONS ( 2 SAMPLES )

Dependant samples
Difference between paired values ( dit =
sci -

yi
* Both populations are
normally distributed

Mean
sample variance
f. mean
sample
sampling distribution paired difference
d- '
th E7= Eiichi d- 12
di Csd ,
-

a-

ton
"
,
=

n
Mean :p -0=14, variance :
i
-

confidence Interval
Cil variance unknown

Cl Ms d- t.tv :-( fda )

:

where V=h I
=
-

,
,

→
If Cl does not include a zero /
negative number sample A sample B
: -
> o

Independent samples

( Ii Ia ) E
( Ni Nz )
±
(I
-

I 2)
-
-

ME + ME

( it variances known Iii ) variance variances

unknown
unequal Ciii ) variance unknown equal variances
-
-

:( M
2E ( )
Cl Na ) ( Ii
on ? On!
- =
Iz ) ± '

J
-

( M
F (
+
Cl th ) ( Ii Ez ) Itv

Sn! t.sn! )
:
=
- -

Pool variance Csp )

sample
:

( 1) SZ
-
im (n .
-

II si t nz -

gpz =

margin of
error
- hit na -

margin of error

Cl ( M ( Tri f- ( I spank
thy ) )
:
2
822 Ma ) Fa ) tv
- = -
±
g.

(
z , +

)
,
t
ni na
where V
-

-
hi tha -

2
where ( Welch Satterthwaite formula ,
v
-

- -

( Sh! ) ( Sn ; )
- '

hi
-

I hz -

* round DOWN
integer
non
Df for
smaller
pudency
→
-

( in Proportions
F.
Film
F" a ph )
Varus ,
-
-

=
+

h 2

f
Cl Cp pal
fi
:

(
fat )
-
. =

72hL
±
LE F"
-

* Cl should NEVER be inte

prefect as a
probability leg :
lecture 9 example -

Cl does NOT show that women have

higher probability of getting a
college
degree than men
, only that women have a HIGHER PROPORTION of college degrees
than men .
LECTURE 10 :
HYPOTHESIS TESTING ( ISAMPLE )

A A. a
Ho :N73 Ho :NZ3 Ho :p =3

Hi : Ps3 Hiilv > 3

Hi :p -1-3

( .
: i ) ( id
l

l
l l
2x O
-

O 2x O
-
ZE ZE
Z
2
Lower tail levy upper tail test two tailteof
-
-
-

step
't :
Hypothesis Eg : left -

tail test
Ho :

Ho :p 73
HI :
Hi : Ps3

a
-
5%
Hepa Teuftfatisfics
-

L :

I -
No METHOD 1 : USING CRITICAL VALUE ( 2C in relation to 2x )
+ eototatiuticlti ) :

corn )

Hypothesis Teotf Means "e L

Ho

↳?f=feufvtaf.to/lsinthecriticalregion
l

Ci ) 6 known -
20.05=1.6449

NI =p .

6×-2=92 Werej Ho ( i. e. t - -
stats 2x for
* Assume normal distribution lower tail test ) -

o
Gx
-

=p
METHOD 2 :P -
VALUE
I -
NINI ,
0-+2 )

I -
No

5%j.
Test Hafiz -
= i

( Tn ) '

i i Ho
( i

↳
l
Z

( iildunknown t stat leg : -21

-

① Area is called the p-value :

probability
NI -
N
falls into the
region
-

that 2 critical
2 S2
GI
=

Ifp value falls into critical

region rej.to
→ →
n
-

( i. e. p - values 2 for left -

tail test )
S
6g =

rn
p -
value = PC 242 Ts )
I -
N ( NI ,
672 ) =p ( za -
z ) For 2 -
tail .
P -
value

I -
No
=
I -

PC 2<2 ) =2[ Phet -

stat ]

Test stat :-(- =

Stfu I 0.97725
=
-

where V -
-
n -
I
=
0.02275

since 0.02275<0.05 →
rej - Ho

Hypothesis -647 Proportions

µp=p
Plt * Assume normal distribution
Gp2 62N P)
-

, =

h since npll -

p ) > b- → CLT →
Torn (
pp ,
Gp2 )
Gp Pll p)
-
=

f -

Po
Test Hat :2= where Po is the proportion test
we
against
-

Po ( I -
Po )

n
Conclusion for Hypothesis Testing :

since the terroristic falls , does not fall in the critical

region
.

We reject ldonotrejecttloata Youignificance level .

We therefore
conclude that there sufficient Iinuufficient evidence #
is
reject
-

the claim that . . . ( Ho ) .

Ho is true Ho is not true

TYPE 11 ERROR
DO NOT ✓
( P)
reject ( I )
x
-

" °
Confidence Level
Probability afnotrej .

a
false Ho

Reject Ho TYPEIERROR
y
( x )
P )
'

( l -

Possibility frej .

Power
a true Ho after

* & vice
Hype -1 error ,
T-type error versa

calculating Type # Error

True mean curve STEP 1 : FROM HYPOTHESISED MEAN GRAPH

-

Find
,
I 172<221 =L
Lansing
' -

* ,
N =

tihf-xcuvihg-xc.to ( En )
2.
-12 a

i
Hlo

W!%,
' '

l l l
l
! Mt THOD FROM ACTUAL MEAN GRAPH
-

2
'

(
,

" ¥-29 ?? I :& ( Area 3. Find P -

PCE IIc µ=µ
#
where
-
I -6449
in
hypothesised curve that

not rejected J
i
was ) N ( N
#
62 )
-

I
,
i

'
,
OR
#
'
i Ic -
N
#

213=1 Find 213 I N 3.

using
'
-

§p
'

2ps
=

! ( %)
own
4. Find P=P( 2>213 ) where 2-1410,1 )
13=1012 >
2131 =p ( I > I ,

I I N*
-

'

plz
'
>
)
=

615N
I

it
i

.
,

µ*=2.g 213
LECTURE 11 : HYPOTHESIS TESTING 12 samples )

Dependant samples
-

( i ) tuhknown
I PD
-

di sci yi 5~NCN-o.co -02 ) Test stat :-(

-
-
-
=

( sci )
-

Sd =
E. (di 412 NJ =
ND
tr
-

' for one fail

( Sd )
-

,
a
n I
( 5) where
-

var = V -
- n -
I
n
IV. E for two
tail
-

Edi
d- =

independent samples
(i ) 6 known

Var ( II -_
Var ( Xi ) t Var ( Xz )

6 , 2
622
t
=

MI M2
f- we
hypothesis @ that there

is no difference (µ ,
-

µ , ,=o
( Fi Fa )
.

( Ni µ ,)
- -
-

Teut ufaf :z=-

6,2 622
1-
hi H2

If set the
hypothesis Ho :P Nato
we as ,
-

Hi :p , -
Nz > o →
claiming that Hi >
Ho

liiltounknowndnneqnal
512
Var ( F) = Var ( Ii ) t Var ( Iz ) Var III ) = 6,2 =

At
512 822
t
=
822
hi h2 Var ( I 2) = 622 =

M2

( Il -
Xz ) -
( M -
Nz )

Test stat :-(

- =
where v
-
- WS
formula ( refer to formula vheef )
512 522
* ROUND DOWN the
Df !
t
hi h2

( iii ) 6 unknown &

equal
( ht -
1) S 't ( h2 1) ga -

gp2
=

hlthz -
2

( Il -
Xz ) -
( M -
Nz )
Test stat :[
-
=
where V -
- hit ha - 2

spank, th )

Teut for Proportions

hip , th 2152 ( total count of
''
successes
"

in both samples )
§= Nith 2

Fi Fa
Teufofaf
-

: 2- -

Ipu -
print that ,
LECTURE 12 : CER

CORRELATION

•
Correlation measurer the LINEAR relationship between two variables
'
- d2 )
Correlation 't causation correlation ( 3rd variable correlation b/w
→
spurious causing a
°

o causation =
correlation

E. Lxi -
Nil Cgi -
Ivy ) sample Holder of X
Population correlation coefficient P ,
-
-

'
E. ( Xi txt 8 ( yi 512 El Xi
2

F)
-
-
-

Sx
-
-

n -

I
( ssxy )
87=1 ( Xi -
Ill 's i -
51
sample correlation coefficient ,
r=
-
sample Holder of 7
'
E. ( Xi -
Tx ) 8 ( yi -
512
Sy El Yi
2
g)
=

(
-

ssxx l l ,
ssyy

=
87=1 ( ki -
E) ( Yi -
J ) h -
I

( n -
l ) Soc
Sy
Ssxy
✓ =

=
COV ( x , y )
ssxx Ssyy
Soc Sy

S ( sci
2

& E)
Ssxx =

!
-

Covariance between X & Y Cov (x , y )

=
n
( ki -
Ingi -

J)
,
,

i -
SSyy= El Yi -
5)
2
-
I

COV ( x when X & Y

independent
°
y ) o are
(r O )
-
-
, -

ssxy =
Ei! , ( sci E) ( -
Yi -

J)
°
Cov . is tied to the unit of x d y

↳
eg weight 1kg

}
: x -

cow . unit : kgcm → cannot be used to compare

y -

height 1cm
t

user -
value to compare instead

( formula allows the unifvihthennm .

& the denom to cancel out - → r -

value
has no
units)
General Guidelines -

>
VERY STRONG CORRELATION COEFFICIENT : r 20.8

>
RELATIVELY STRONG CORRELATION COEFFICIENT : 0.GE r E -
0.7

>
MILD 1 RELATIVELY WEAK CORRELATION COEFFICIENT : 0.5 Er E O - G

>
WEAK CORRELATION COEFFICIENT : 0.4 E RE O - 5

VERY WEAK
Dependent
> : REO 3

Independent
-

Variable Variable

REGRESSION :
how much y wills
given a sin x
X Y
simple linear
regression motel :
f- Pot Pint E
Exogenous E dogenous
variable variable
where E is the random term E NCO
-
error ,
-

, Ge I

line
simple linear
regression equation I Bet fiflinel Predictive Modelling light least square :
-
-

Y bot bix ( aka y Muto )

- -
- -

where bo :
Y intersect 15 bi II Ei =
Difference between raw d
predicted value of
-
-

SS " y
bi :
gradient of the dope ( ssxx ) ↳ when
you square all
the Ei drum them
together ,

the line which the minimum E. Eia will be the

gives
beat
fit line
-
.
Auuumptions for Regression Analysis :

i .
True relation vhip form is linear Cy is a linear function of X t E

ii. Ei is
independent of X

iii. Ei is random variable with &

constant
a 2
mean o variance Ge
-
-

iv. E i are not correlated with one another

Sum of squared ( AEI

Ji )
' '
error
usyy ) E (
yi / Ee i
> :
-

of square regression Iss R ) /

SS
Model E Cgi ji
.
>
sum : -

↳ difference
between predicted value y I mean of y ( ONLY raw
look @ predicted line )
Sum
of square to -14 ( SST ) SSE Ssk E Cy J )
'
> : t = . -

y & J
↳ difference between
>

Point estimate for sea : Se

'
=

YE ,
.
where se is the Elandavg error of the estimates

R (
-

coefficient of determination ) :
portion of total variations of 7 that is
explained by X

Ji z )
'
SS R E ( -

SST Ss Rt Sse 8 ( '

if I ( yi
'

yi E
Cgi g) E
Ji )
-
t
= - -
= -

JST SSE -

R2
2
SST = r
SSE
for simple
= l -

S ST regression

Le auf square
estimates
Cov ( x , y )

87=1 ( sci -
E ) l Yi J ) -
n 'T E 7- I ( sci -

El Cudi g )-

rs , sy sy
✓= =
= =
r
(
Sy
Soc S"
)
n÷ E (
n -
I sci 2
Sx
-
-
E )
,

significance Teufel inferences about the

dope

step 1 :
Hypothesis
Ho : 13 ,
=
0

Ht :
Pi to

step Decision
2 :
rule
p
-
value s l s a

con duoion
steps :