Linea Kegresdorv

v a n i a b l e s -

elati ona
hip among
dentiyto
Ldenti'y
them
make
PMcditions

Bet PX +
Y-
dependent vuable
explan allory vauabls

1egne% ion
Sumpll u n e a s
me explanalory vania
abbl
lee

+ BXt e
Y= Bo
lune in seatteu
add best
besl gt
n excel
ue can
pent, then elelng
data ptnt,
clicking on any
plot by
a d a trendliKe".
the &um
en o nmun umise
Valuu Bo. B cho
99uare

undu e a s t s9uase:
Reast 39uaue:
Aumplëon

uos a undepenclenE X , X2 . Xn
X
ECet) = o
expeeud valuu lo 1e
Du have
vauanu Van (eI) = o 2
Au os have sanue

ae
uncomelald Corlei, Ej)= oo

he enors

ane valid random os)

t u s t two not be
Vauianu blw roaod
Tange dala may
hours
vaianu blu sellung
pua o Cmall

Came :
targe hous u)
and

(neathes)
Can 6e
can 6e
Comelald un casd fove castin9
uer
 ANDVATable

Vaniane
(Analycik

SS MS F P-valu e
df
Kegreusion
Riuudual

Ttal

degre@s o udem
f
SS
sum c9uas
MS SS
MS mean df
explained vauatibn
Reg ression S
7otal Ss un expaun ed vaiation
Eror SsS

SSR
Evplaned vauabelety 2
R2 tal vauability
sST

ditevmenalin o R squaned
coehfeeteut of
explaualorj Powe

R cquane sampe
or smple g r t b n : Comeldtion

Correlalion mcaures stength a Diuear elationslip

vauiabls

Totas SST

Ji pndicled vallu
Explained : sR =
2-7) comespondugox
1
A - Bot B, Xi
bnr et Yi Y
2
Unexplaned: Sse =

R SSR

SST
 Matng Preollfous
or expe cluol
awerage
Jndiudual values
Voluu

ELYIX=*] is the xpeilaliou oYgivenX=x.

+e, Cerretponds
to igt
So
Se, y grRsfen line

Multiple regrossfon:
PXE + E
X.
Y= Bo P+

7Mulhictlineauby possibly edundaney amou

explanale vautables
ingh value
ihat buu slepe may 6e Fuo
-value!> uggest
uhn oth er Vaiable aw ndudesl - ich
which maans
u e can wmo ve ihat vauable ba u medek

f value F-valuuz MSR

MSE

Adjutd R :

Extra Vauablus Can entehou wth proper umation

lcpe
L a l o tuallA nouas R2

vaiablu
Jo penaluue ers
df

adjusts R sSEn-K-1)
sST7n-D
In conbast R SSR sST-SSE SSE
SST sS sST
 ds cale goital vaufabl olumm
dummy vauablus

uACol

locatrons Eatt meadouo

Levitous

EM
enodun
L.T.
Ieip
dunmy vaiabls
au

durnb Calegoies (m-)

used

inear grsion tn machune leasning

leosning

"raingung et|

ea-dng algo
hypcthera predictd Y
tunetton

ho()te aradiet
Tero
condilion

Cost unciu J,9): [h)-]

muasues uaus olese predictieus au to adual y elus
as to minimue T(,e.)
L9 ie choseu sueh

fuuebion: >Alao ued gradeut dutt

Miimisn Sow

Lclaut uth o,,'j

tt nduu T[eo, P,) util
ksp ohang
we a c
Do,
munmum
 o undid .(Mnnr
D J ) , (g)

Can be d
los unclion
urelitn
euor
udual
Vtung
as Leost 1qua
L1e,0) Z i - -e'ni)*
2 Mean 89uare
--0x)
)1eo,e)
nhrl Roet muon
uanlo
Le, --oTe)
n
ou tuneton ttep,0: 2 (-0-F
DDehine i
Oplumizaion oblem

oplinzalion Preblen
Solung B- bo- 6'x

e
es

W
)y-xa) (y-x)