0% found this document useful (0 votes)
9 views24 pagesML 1
The document discusses various statistical concepts including maximum likelihood estimation (MLE), Bayesian estimation, and probability distributions such as binomial and Poisson. It emphasizes the importance of likelihood functions and the derivation of estimators based on observed data. Additionally, it touches on the application of these concepts in estimating parameters for different distributions.
Uploaded by
Akshay SCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
0% found this document useful (0 votes)
9 views24 pagesML 1
The document discusses various statistical concepts including maximum likelihood estimation (MLE), Bayesian estimation, and probability distributions such as binomial and Poisson. It emphasizes the importance of likelihood functions and the derivation of estimators based on observed data. Additionally, it touches on the application of these concepts in estimating parameters for different distributions.
Uploaded by
Akshay SCopyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF or read online on Scribd
/ 24
i PScry
=e) a es
I eema 7 Louse aan i Pie
ee oo
Uniform PEL Pe
| Ecponerctral
enaal bot theorem > Ohl mbenyy
| distributions ane __considincoeh __guysston |
[o__ estimate _Paxometer
| / k
_, Binmial Hise) =(") KP)
Poisson POek) -— eé Aak Kk anival
1 => means
k 22 nD
te
Unit erin #() = | oc € Larb j
|
|
Boul PKS) = pPCI-p) nee
Nozma | yi FG ye = =f
ER
api ae ==. ee
Vaio 2#——_——
7 80 LS men
> apsara
Ozte:__-_--__
FR) =e ~
“Espana
Moximnumn Kikeliby ook Estimation
PBIA) iw Bayes Therm
cae a6 assump tiore
Ret Xiy9X) XZ. Xn be Samples
Finds -the distnibution trom which these
camples come From (4% assume ) depending on be haview)
Finds rts probability function
| Finds the —_unknduy amelie in
|_porsba bs hy function throug h Maximum
dikeliboodd Eetimations
io ings those set oF porametie —foy the
assume probability ditetribution we thak
ives man kikelihoool gr the otourrcniey, TF
=bhe Somplis in thet distr bukron
chops
t | Collet Detar [Xi X25 ++ Xn J
|| Assume probability distyibatron
3. Famulate the Zeina Funetiory
aA
pe
maviMi3, this fr le fire tfase
Wi pavimetews for which the “Fi tae Mae Val Likelihood’ Functions sf slandanae Ban
TO EBay
| —
ani Form Lal events equally Pikely ) ~~
ev can be dixvete oY Continuous |
—_ a
= |
To Feleteoe to isso
= | Pla a speciic™ “opla blw suc(essive —~"
= _| wnt " “gecur rene)
en = EO heen
= = iM oe Xe Xn)
a Ww ’ 5
—— =n ¢
ak _ 1 log(L(A)) = 1 fog —n[SXi)
a LOY = SX
Pe __ av hu
a tog (Ltm))~=0 |
a a oe —
an -
at > [r= 7 = exparnunkt oh
4 = et a
Hy oo
wok = = _ apsara
~ Qaussian
pa f = I —é 2a2
2 pawametens — by x LAsume | est while
imotirg ethan )
for all _unknauin distvibutran map _*t to
Gaussian because ot — Cerkwal Lime Theaver
_ (xP / — Kyo eae
(= | eg 2 | ¢é
ee
ae Var
(S&F)
yO
ne dag CLC) __Log (Gro ;
ail “He Aps
a ie
is P(e DOA. cS
|
er ee oc oS
aes See eee) ——4
_- ou
= a
_——_| cm
i | Feet uiatr =o ——t
co nse —
: i 2X;—= LM —
~ ] a = —|
H :
i ee eee
1 ———A-} Neverat Dicvubaton
A apsara
ste:
ey a
log (LCE) = toy [#P fog (2AT2)
ZL
= 7 (5 (x=2?)
Loe
yhg(I()= 0 1 x 20 2%
3 4 LR
oe
— 3 6py =2—
Zz a
= A= si -Ky?
Lee
Nb Jteg (LO) =
oO
> Ss 2 =e
Le oF
= go? 3(ki=u)* ZR
a
ge No. = | 7
Obs
PSQy
ak
— D. —
a — =i. —
a i (xP) Von weg
__ 1 0 /
a Ticahoe OWT pe
_ _ Mawmun Au keush MILE ge vey
= | cnamay L(e) or
_— 7 J 6 Si
— &
Mevmuin ow postenen Estimation (mA
ae | by Says Thess 9
_— J ZF
HH [ elxT = pxle) ole)
a7 | POC)
eT | P(x) om need to be Gonsicltna
re? | SIN ‘tic constan t
ra ___ Plelx) =P (xje) ee)
re i—Io_MAP, ging wrth MLE
al (prions alse fongideveds
HT | oy e(a)|
aT _Uistai bution ef eo | . apsara
Pe are 2 —
| : —_
a), Davive” Derive the Mab estimation fay a srt oF
fo K | data XpXy Xn tun ing (tom _&
’ assaf deste ution
m= SE (mean ) _
, {iVven pruiow oF H Follows ow gaussian.
| lla ibutia mean Wand Vowiehce
1 I || Xp oX%yy-- - Xn 2
J 4) part= Pie
} | V20 o>
7 A (xi= A) .
] L(#) =(/_1_\e < ert re
i 4 F
| —b—&y
3. Prviosy : Zp
| pte) =
| oF p2 _
4: Posters o _
| p(uix) = Le) ~ PCR)
J ae =F rf
| i eS L e—ip*
a as - _x ——
4 ae =p?
a = ‘Page No.:_ 22 = 5) | ‘de CPC mIX xy)
— 7 pS 506 sh
ales a) ape
—$__
os le po? =
Il
|
i
il
|
|
|
|
i | oe
I 6). ‘Taking dewivatiw oS
7 Ty ag PCR)
re OY -
7 =F (2Xxi=B)) = Z(H)
he £02 £p2
=e) ae)
Be 2 as
/ 9 r
a! 7 quake to O
He PRR) G2) - apsara
Date:
eFax ee |
een
MLE > avgmax L( 6
4 z(6)
MAP > agen are PCGTX )
&
Ky P(e)
A) Sup pose ou me given a col ant youre
NG sure whether “+ is faiy 6% biased -
You flip the bins _[o times _ te TH
gad __3T Assume cach _-flip _!s hnclepe naire -
_|Then timate _prabobihly fF getting Lilt
A A, XX pe Xn
2) Find distribytion
7 0 anh | => Berogalli
PCH) net Binomial (We ave ret
aki b Find plano oF suc)
pm=f-=—0*(7= 6)! =Kk ony |
TAL! 3 = 3. =0
o re
W1-8) = 36 SEE
ere
ae aad ;
st
ae apsara
Date:___-__
You ave. still Flipping that same coin -
YYou__obsevvel 1H “and ST. Bat ow You
hove a prioy belief that the asin 16 slight
biasch towavde heal: Th pliee io leat
a peta — distibutian “ith povometerS
a= 4 and pat
Estimate -the probobilrty + getting head:
Assume Beta distribetion ~appraxi motely
ae Yay!
Use MAP Sine pricy |S giver
! =
Xi x. _-. KO
| ! 6
2-| £ikehhoook — Besnouf]! ;
L= pott = of (-6@)I*
fale = p distributia
P(g) = o°(/-6)
3. Pos Levioy
—__ ss a
p(elx) = L(e) PCe)
- 3 = (1-6)? . 07G9)
= glo G04 | Lo PCOLX oa
D fog (10 ) 44 ta
|
I to]
& Cég(CIx)) zi
| |
rere hh
Ss
l—@
fo
(guating
z ZT
0
lo(i-e)
4G
—— fi
& 10
=
= O- 774
mice cel, cl tt
+4
ee
=
5
Suppose
yu axe
anal time betwear
Machine
dotluyes which
mode!
ret ia |
7)
Aist ys bub’on
obser the
|
|
FGI
following
4,4) 3, I » 4,2
the.
—Aale__parvometey _i
we
4
A
be
“y-
Page Noe
\ apsara
| MLE
| pat = Renn “n
Date:
LO) = pe we ee
= é
Gia 1270
Log (LO) = 5 Rlog(r) 12H
dh toa L(®)) = 5 12
Pe wm
“Ui eS = O41
12 _
You aye still analyyng the same data:
‘| Bat now __basech _on St
experience
“ypu believe thot _-failux yato howe!
Abe amund 0 — ae u_model
this belt dyistyibuliar
ae at
| burt porameles k= 3 a pak
a Asscne — YY distribatloge
im Xal g PX
vem ae en
as) = EO ey
ES
ur tt
oe i
Ed (hal abd) uaa) = 187
aa a : 1 1
aa | ah mv
Tid i v= = 034
| Pe i (3 ——
4 8) You obspwe follouring data pati
PB i am _ Sov JS
Ho 12,15 518, )80, 714
_~ an. Hey Gre dygwn__indupnds
_puvogl distr bub or _-
hth mon prank Varun &
== Te 4
eng apsarc
= LQ a —_
i = rm Qe oetey =
= Pulp) Pole) pC Kater)
=f oie: jel ae)
[ =
deg (Lh \)
Seay ft VES
2 -\War02!)
a
L{é (2 So + G8 P+
LFZ
go= + =)
- tog (L@) 7 a
{ZL aa (12-1 TSEB =F +)
ae
fe =—_—__40=p = tau} ei I \ \ \ \ \ \ \
14
\
\
Pea
Wh
\
!
hh
|
|
i
I
IL
= Pee Xoo)
(2)
P(xilo) PX Jo) L POX n lem
5 I y
=/_! Zoek
ce
(WU 2Aa77
oO eq ((#))
= =F Txgx Ff
a thge—_, go
ee A Ce eae
o3* eee pga No" apsara
@|| You aye studying +e no sf _emarle
Yeaived po ~ Ow a 6 houaw
windiu/ you Yack -the follwing wunto
3, 4,4, 0,16
Assum (aunts fallow paisso?
Ctimale yak _payannhey Jv”
&|
beliwe
Now payed past behaviaur pe
9 jaa y__and
pegple ipreli pltell see cnoeh
friuz ated YsINgG Y.
disbabdisn wath, permet a3 Bald
- : Ve
J Pobre = @€ AL _
LH }