[ | Pow T EstimaTion—_ Ctabickical 3 E “ie pines of makin inferences about pools fon bag on samples taken from ik is aalled statistical Inference °F ima tly lnduchve io inferential statistics gkatistical inference ts drawn Prem Inature - Because, condusors ave based op the ample the popes Gon abou the —doknown ag pects of population - problers of stalistical inference f theiy solutions are enaialy kinfey wo dyed - (i) Estimation of parameters (2) “Testing of stabstical hypothesis Heo of atimation was proposed by pk RA Fithey, a fivst™ about (980 __ [Estimation theox Te edie of ceming be a conclovion vegasting the Nalves oF unknown pasamerers based on the information port by ample 1S frown a3 the podlem of e@timation * sre estimation an be made in two ways? () Pink estimation 2) Interval estimation - [Pointe estimation IF Free the obsewatlong in a Sample »4 Siogle vibe Ys calculated a2 an astymate of unknown parameter . The procedure nefevved to 05 pointy estimeton & we yefey de the value of stabishics a8 poink estimate . oat | Fyrif we se as avalue of < estimabe theof & poptlation | we pe __|perinable _ ‘4 Jer_of a good exthinalpx - Accosting ty RA. Fisher , flowing a1¢ ; 79 the dairabfe. prperths of a ged esbimylor - P ™ (y Unbiexe| ncas (2) Consis Fen cy } Efficieny fa) SofFiciency | (0 Junbiasedness: consifey a popelation with pat Pc} F(0) whore ek unknown §payameley . ick T be om ubnaboy tr eatmate 6. . en, F 7% suid bo be enbiged cstimatyy of & , if F(H~6 OST , sarple mean 7 an apbised @tinaloy sf yopulation mean us] Let, x). be 4 santo sempje of sxe oO taken from the prlition ath mean fs Let = be the medn @) F =7tGh- +2, qve point alimate o ving a x n E(t) = = (ees) a, =le(ayet.. +(x) | | ne of | y eal —1 QB] check whether Sample Vatiane % unbiased foF pylttion | Neiance Let 4,%,.—m, be sample of Ze n taken Prom de Population with variance o? . Lek & be sample mean § €be Farmple variane . ae | @ sta 3 (a- zy " le a B= e(L &-Z)) =) “Fels Ga] “EC H pongey Sigs - te el] TELE [Gi 26 Weyer eh | eh falda Pat Zaey ="eo PaaS pr ee yy our ae % a wbiete| ex timaboy of yal ation anil of examples ave Ken From normal qopelation . LE Xjp%ap--%py YS Tandem Semple carth From novm al jSS Lt tn N(Po) ST TH Fey Te unbiased ef estintoy of pet pal CE 327) | ~ (Reel) —e | faves VO) = E(x") - (EQ) G2) = Vbd + EO? Oe Flt) = AE Z (vp) Heo)" ] 0 Soy alte (el | Sit | —— | “-T Te an anblaed esbimaboy of el A Lek p be & frofevtnn of diective item prediced by ¢ Machine = het theve be n produced by @ machine gre examined f a | Inv x’ is defined a3 follows , x= (1 if MH ikem & defective i st if T=2 0? then WY unbiased estimabny fox P iy) 10) 3% onbiaged eatimaboy for iz es n(9= 1), | hate, Yv X Ae byibuted @ Fallows , ees EP [Py _I DE (%) =% E(v) =% (2%) =KEF(i) | = hep ! a ; vpaoabelait oR voor, dally ® [E()=0xg) + «01)) = f]\_Xv8(np) - 1 T=2 5,0 8() p Eq=op NO =r E (Ys) = WE(1) =YorF =i =—— i == CD wee do get Tew) ae Shen PCH) oh (te) y ncey En} = ae une Fd -e(nyh 2 abet or a asp Dt PHS = ta 5 - AGP f) = Po mt Pp i Find uobiesed — extimatoy of e4, wheve 4 1B faremetes of foluu Nistybution - ’ Fiom & population ith tH oo - er 5 X30,Bi0 Rear fia vendon Gerple of sixe 26%,c:T OM of observa anbiased for & akimates GT 2x % unbiased fof ® > E Pt ey = wh eve ore : tal es ‘An independent’ sample of Haze 9 % ken From =f peta | with mean pf Varvience oe thm eT ae y 1 eLek x1) €X,,—%n be & random Sample - Ten, le-us dobing a Punction P(A) ~- Fy) = + rahen 220 0 j ow EL arya to = (x Fe=8) + -FCa)) Petro ot] = &140 aA g is the anbiate) estimatoy of a) © Pwe) pee Leponntid etibsion] 40 10 xR, ave the Fanfom—kamr-pte observations of random sample . BSN Ae = x4 Yo =m ee 717 3,°) = 4 Yo _ Ya a! = 48a? ”) — = ia ery? ‘ "2 opeew since they x (4, ) E(% are run from Seme e £ population dictsiboften So, (x?) = Sa! tod they axe _ identical . t oI = fx 1. aie ° “ a Py _Ky fat o7tds -£.] ot fae a Ode (s* : FA ° ob ei ur\ Ye Seoahek 4: = 1 %a. e We (me) = L aie hy -9? er ec et ag Ea (mate J = VWre ee© (a). €(39") ree % Se ce or unbiaic) timaby of © Pa) = »%tm—_ —4 Xv U (0/8) Ee?) = 2e(7) x 0 ~ Cre) = 28g 4) =22% = [exh = 248 =04] = SNM ‘ QD ok R an —unbised esfnabor of & where— kN © An—independent sasnpte—of—age nts taken—trom—pe portion meen pe tiamce j ct Snr a a Fy = €(tay CD Efe — Catt ti = om LEC tl +s + = = (fr ail f = ore (% iret Uttam “fa (y-2)=e ran ye = 2640) a cca ajip TH an anbined estima box of e , the 8.7 TV fie ye biased extimby of o PIR. wre Mo) 40 Pr) tal N(T) = ECT eqn)” =E(r-e) go FE(T~ 2Te+0°) £0 ——> ECT) ~ av eqype® 0 ai E(1?) —207 467 #0 pal F(T ee TW biaed stinaby of 6 Now, NIT) #0 YWT) = (RF ECT)” #0 = E(t) ~ E(t) fo + o- Aa #0 me Fi I hae” utes eqeare on E(m) +e “MG Ss baed atinetoy of Se. 4ST Sample mean TW wnbiazed cstimaby of -E &y the _ distribytien Pose) = rial a= Ga. —, oLeci 0 2 ant £(z) Era a eb ee a) E(n) =En. e(\-9) t=1 = 0+ 20(\-6) +3e(\-e)t- -- SR ke oe ee ae= of-(-olf = ole =o I | eeye tdey Bhs, 0 -L2% 0 i =% “ & B% anblaged estimeby °F Ye | iyo tn sangom semple peiwon Uebribotion with parameter ae lestimatoy fer 4 8) Find or snbigned eatimaby of gevemebes 6 Ustibublon V@ Ufe 2] bose on the sangom jo. obtain alo the - Variance ot ; \ ee ez) = EE) =t+£(2%) 9 Spe (ti) Prom @ popale fon Polly Sayer any 8 unbiased of anifosm Sample °F Fue E(2\) = & =! 4% = % n = & ~> ut o abiael atime: €22) = 2, EQz)=. - 25 B an anbisred estimates of & | v6) = 6) | ba 4N (HSA) =e 1 Ven) reer ee== F i= a 1 . q ‘Ge, 0) = ales © jrmdeed, T oe é 2% anbisted fox le 4 (a) © TEE ae onbined For & y Find 4 Such that @ 47 HI-ANE Jhag leant variance Yani) 1 ao Az) = Efe = freade(12) E(t) = e(r,)-0 TAT +O-4im OO = 4 ey + (a4) Ett =e +(1-4)0 = de + 8-40 So oy TW unbiased fy 9 “O) = VT + hay) = ANC Ot" vm) M0 => 2A) t20-Hea-l VE) = 0 ae = UT - 2 VCR) + 24.V(%) =0 TANG) + 24 vem eek 24 V(t) > 240n) + va) J =2v(ny 7 Aven) + v(ty] = V(t) 4 = May NT) +v(t) 7 Ane 2G T+ 2v(h) 19 d4*in estima OY Th =T{ x, ya9-- [sample of Sze no T% aad to be of V0) ,8€Q, the parameker Fpae » IF 7 Je) in probability - uw, 'F Top ye) 9 na In othe words , Th iy wnaithent eahmabry of | for every & Jo yhro , there ensh @ pre Mt o2m(E,7) such that , PEI To- Hoje ef— a2 N—F od oR PE(T— We) 2431-4, fy osm , where tm % teme Va Jatge value of Y | ahich Agenting we) ‘byarlance _psoperty of. psig estimator rk © warsic bet estimator of Ie = YO) un tinuow. Function ; esimalry of Y(V(6)) - Proof 3 Boe) 1, is a wnngtenb es matry of Ho) ¥ eto) P10 © fee NAm(e,e 4) such tht ! Pilt-%o)] Ze FAl- Hy for nam(ety Since -P{-) is ite euntion, theo fey evey there exits a punitive number such Yo7 Maye °(B) : a @ ) (8) 4 pra) é From (2)¢ @) ; =P i\ Vin) = Yoraylee 5 299 |[T —Me)|4 3 PAY ty) -¥ Oo) | 4E4 21-1, > fox every nim Heme prove: Softident” tandition for consasteny) Lek m5 be a sequence of estimalry gach thal foy ali ee8 , ® €(T,) —3 Ye) og ne Py: 40 ate Then) Th 1s ansistenl” estimator of Ye) Proofs We have tp prove that T, $3 Ve) % 978 1 PE [To We) 4ey 41-y Am (EN) f where @¢€ h aye aybitvavi ly small poitive numbers and m is tome layge value. Apply iy°9 chebyshe v4 a b Th, © 0 PET tm 28 }-sC) Weve, =I) = Ih- E(t) + E(t) “ale now, oe F(%)| £5 9 |nSt|e 64 cS 2 thle iyHence » ‘SOP pose 16 ASB, wu ASB =) Hale P(e) ow i, le ge ISITE) 2 2 She riIh- ~V(ol|2 : &) PEM-Ve)z srle(ny- Yo} Ipgir - “ema yy, rmé Now, we aie given thal’ ° EC) —— ie a nee N(T) ——y 0 o nyo q Hence, Sy every S10 , there erlgts a gositve inkeges | 04N,(8,)) such thal \e(m-ve)yzs, ¥n4.(8) — © jAlo, V(T,) —7 0 os nae) weve N(%h)_ 2 y, ¥h40,(y) ———( a 1 whee h ® aybitvavily very small positive number Substitute ©¢€ (D in (5) 5 we get - PIT) — Ve) | 2 $+ $43 i-q ¥ » 4m (Son) FEIT Teple et di-q, where 94m Shere my = maxi of Mo § Coax (nond y] ¢ €= $48, ay ht Ye) 8 nw <) T te wnsisbent — estima hy of 99) . Hence the pe of0 Matas =p a i Fol tite om | = _apbiased a CM}ib er) is ampli em ais popula Fon % suntan extimgtry oF > aga @ e~* sxcaochys istyibu tion Rah apie mean but mee Te a wrwirbent exlimatoy of sample mean {9)'s los) Sampling From N(We) , we know thal @ , . robe : Ton x)= § Wa=% Thus | 3 nv ob E(x) =P 4 q=0 [e,—7o | a 0 X is a wonsiefenl estienator of mean po - | ost) iv ty Weve Caachy distsibublon with pavamele {ais @, Xvc(P)!) @) Me) =e HOS ay ooh pat E fs | w+ (rp) , to be eust Tren , Prnean “ab” he Ustrbution emomed bo 9 ak me : : Rif Same mean TW taken w W- estimatny then F(z) = + rep 5 Hep oO [sine disk atb otlon of € is seme er ip couchy eqn We docsn’t | with tor® Fg Ww not & unsistent extimaloy of yo coh Jekvbulion - we know ay » caudh aivGabution % rymmebsic abou we— population MUN Ge Cenean=median = Th Te enugh fo say thal sample modian My Ye unbiged for popale tien median f&. , E(MA) =f - sampling distribution of modian 1S pox ge 9) poymal and te ie given by y aogympirtically F(a) = X ep canF(z-Y) J where fF & the median Rk we exh S (ee ne —_ ) rh Yat if ee Herve ne = maian oxfinake af 0 aH) ee Loretta "Phere Hence, Prom @ re! ; ei of sarnpling Ajekribubin of median Bia) = | ( ) hop™ oaioate of ore? C popstien - ne =0 wg n7% = sae ‘ we wnclje thay > FO caychy'S isteibaben on & beinoulli pi we value di ~i) 2 iF arter- Fa we sangon observation Rerakel Giking eer Yates with _pieb> with pb hp 7 ST 25 (\~ S81) > pan a ail ssE(X) = ecm) + (rv) & - ay « <|| 3 = ] mere, ERV=P 4 V(F)=0 B nm \ AS consis ben’ estimaby of Pp» au, 2%, 2) = xi-7) W pelycomial ; a 0 i being ny) } polynomial Hs i content fo py Visca porty ofibe — 4) 61 sample, mean X ie unsebenl box pepstatin * mean ' a)| st Femple Vayiance — consiztent estima boy of pequletion watiance the. we OF N(Eoe) n be the no. of bernoulli trials» X 1% the noe os 6) Lek sucess In seve of 9 ria with prob- of suuess | p Pov eah trial. Then, ST x, Ty consistent estimator | oF | (| s-T Sample median 1 eonsirtentt abimator of population mea of noimal gopelation (5) VF 1 iS consistent otimatoy of © , then F245 2 iy dlgo a cersigtent estimator of fa . \ eam ple—mean Oe q cquchy’é kick: populations amp 10 We ¢ know shat c= EN(X)= %, a nan 7 E(2)= fe € W@=0 ee Fe + nclting estimay of f meen epulahion ® 0g noe) Ef) =o ¢ ww)=0X Bini p) E(x) =op V(x)=9 p(p-1) E(%)= yp E() =te xop 7 eh NG) = 58) = pale) = Me) a ao, EO%) =) q N("%0)= oO [a oa] 1, is wratant es haley oF > : XUN (yo) In Normal Aistsibublen , Mean = median = mode = fh em) =H Medan stdinate fe es Na Viance , Vimy)> , ; Given, T te the v 1 Alo, T7 ipo polynomial in T. The Oe Wt, We continuous function in T - va “TT We the wositheb estimatey oF oa then — property of ansictaoh estimatoy ,T7 & 4 oF sf ev:ir Ffeuent etimby Fra one u Eter—f thee ave move than consistent estimator all enisF seme bine - Foreg yin cempliny Fram Nfl) “hen © koown 9 sample mean % Gf Sample median My ne both aunb ios an} nsigtenk estima oy of - So fhae © wome — Fur they cnteria which will 2able us ee & bet estimators beruren them - a, wbich %% based Nf oe Sampling dis tibulon ef thee eatin « boy 8 | which 1% known 98 effideny - @, TET, we two ex 0% — cesttin pavamebey & - IF we have WT) 2v(t, for all n ) then | 1% move eFfideo than T, . consttent linabors Eq? Pe J* Soppse T= — sample medo [ eomtne T,= My — eample median vp = n N(My) = = | 2 1 N22 | a) a Us qin ia we, CH) 2 (MY ; ee ie , more ef fiden™ esbinaloy than the | | semple roedian | prtaweter, there @ni86 one whose sampling Wi his than thet of any Buch a iif Tw the moe’ vavute \, ¢ G ty the they &Kimato then the ef fideny of Th e defin VN "5 No obviously ca exceed unity * | Ey? EFfiden y of T in the above | E= \(z) cose 5 Na) =0/n [e ts aloap toy ee 4 oh =2 XK ol is eee Oo A random — Simple Hy X29 Pre — neoiroa piston with — anfroadn mean Allowing eg imatort db estimate # fH ee E i bh = X 4% ae sees T | ts = 2y+% AAX | & ee ens onbiae4 where A % Boch thaY *3 ; 4 fiystate § f ptpring A. ware Hid aobiaed 2 : estimator beak uhich 8 Bek among reason, be by = 28K 3(Guat) ap seta) reCs) F486) 2p 3 ieee =a 3} tAp = 5h 4 =0 = 1-0 (eed 0D) E(k) = E(x pata teytts) > ~ LEG) +e (4) + E(t) HF) +F(s)] =_ 5 a ao | --*, 18 anbiased esfimaby fox t- | Eth) = E(M tt +%3) = OH) +P = e +f ot - fe | <4 TH nok — aobiaged —eatinaber ef FE - ME) = N Ge (atte tty + a tt5]] 7 ee eG tate taytas) = te POO Nan ee MSal Be [R# ve independent ale 25 \ ols, | N(fa) 2X\4 Xt 0-X3] KG A) =dvid « hs 3 =4 \V (xy + VC) J or eed = 50 4 N(h) ~ the &, Ww bet leas Bie > %2 % % populétion ith — mean ae the eximatois «wed «vo es images mean value fl sanddom sampl & of nxe 3 From w Ww & vaviance oe. teh 144 wheve 1, = X, £X,-Xy | Ve = 2%, 43457 4X T= (AX + Xat X5) i) Ave Tat aatiliel estimator % W ting the vae of A such that Ts % estimgtys for fp: (4) with this valu of 4 > 1B Ts (ND Which i bet estimator 2 anbiaed wmiztent estimator 2 sae AX, ~ Xq ET) = BGreGs)‘i Shana See =F tx. %) =H = E(X) + Ex) Ep) =e = TP Sh ee = 7 &® vobiasef etfmatoy of fp i b= Sky 4sy,- 4%, Fla) = © (any 3x5 4%s) = 2E(u) T3E(K) - 4 E(x) a Te Ts unbiased eafingtoy oF P | ee T= AX, tH & pak = Fp cy p | SEE Got ebay rep] =f ai et aaa dp tap = 8p | ae eae 4 = SS a) Te = MT a TK 3 a ’ Fample mean Ys conaittenk etn; } be FE NDX AN Dd - v0 a5 J Fete oa NOE) = VaR + 3 a tex) = AN(Ki) + 4 VK) - 16 V(X) Fat ot = 2467 NESEY (Gar OES) 3 B) 5 Dey + vou + veal] N(G) “Ty te the best estimator inimum varviance -1 Minimum variance unbiased astimatoy [uve] a ee Ie a wtutislic T= Teayrtr--ts) bared on a sample of size ‘yh’ ie auch thal > CT ys unbiased fox Ye) fev all ce OF ji) jk he the Smallest variance among the lass of shape all unbiased —estimatovs §— °F 98). Then Tie called Teo T ig called —aninimum Variance — sabiazed estimatoy “es [MWvE] of 90). ly they wots, T & mvve of Ye)» ie E(t) =V(9) tee® ¢ M1) £vq') Py deed, shever ic any ofhey unbiwed elimatry of (5) Theorew An MVE je -anique the Senge Shak 16 7 4% cre MVE fey We), then T=, almort surely Pre fe Here, wee iver thay (we onune Ts Em) =E(G) = We) ae % Mve) | \G)= hye ¥eed i New, ansided @ new extn gto petty 2 Ws ries 2 SNE [e+] = Efe tet = £(e)+%e] = 20)= Ye) int Ab, Nit) = NT yO Mes) = NGF) 4 S=cov(xg) =i {cw + V(Th) +2 Ov(T wy] [ aS Sa FfrG - ove f aPLVEmy TY (ryt Fate ey] -t_[ev(ttapvigy] Es by @O| Ve Liv) b+e sy aheve § is the ays\ pear 00's wefidat of wnelaton b/w 14 | since, T, is Malt vue ,- VT) 3 v(t) (yn) Fs VEry mst a - 2 | se | Sines ve know that |s| Z{ |. we most have Poy @, TAT mack have ¢ Uneay we wefahon of the form T= #4 pT, ——— ) | © here “gp ae the cnskanty which ave independent of the sample Ky AyyAge > Xq P bebe lh may epend on @. wu, ae ve a) have <= (9) g pe pe) x Takin expectation on bth sda of @ , we have F(T) = E («+ bt) 1@) = «+ pie) ——— @ [+ by 04—= " ote, vey =t yx the wepfiden oF regenin of 145 postive E ahs x20) Ey Oy ~'- Subsfirote the value w 4 B in @ , we get ae - MVUE K Uniqae : 'F TET. ave two unbined eatinalnys of Vo) havin the same wriante 4 § ig the correlation wefidat Y Ahem - Then 8.7 SEY, of exh estimaty, ony} lel T be NUE ef Y[o)- Then , nce NT) = V(T) » then ef ficken y ‘e’” of the estimators pwhere e@ ig the efhideny | is gen by» ast) Sit) mii ‘(ae hil Conidet anothey liiaed estimatny of %) “a 3 entry) | te, Wy = NEL ental beerly Lycry eect Seg a =1 \ | Ne 7 oe = Ar) lise j “o[! | \ vey = ve) 191 2e@ Gime, VET) i minimum , .- V(T) 4 \(v) N(w) (49) Ay 2¢e = ©) = See 2e / = gy dze-t —S— Sof iden y An stray i said be all the inPosmations be supfidenr fos w in he arametex , Ye eobains | sam ple segar ting | In thet wot, tf T= T (ty>%ar-> e bael of @ sample vith density (p46) F@9) = given J} the pavameley - %) Te an ap Healer jf yoramete™ €,t29 > FN of Prom he population cenditiondl — distviba lion of Ri eau- Then , T Ye vob Fide ok eatimatoy size 9 goth thay the Ts independent fF © doy & * £q* bell Ryton be & andom Bam ple From beynovllt population with — parameter v8 Neal : wv, oS 5 phe. rea i jo 3 |Pe ee a! ee Then, Reb T= T (a my-~-%9) = eat 1% = én, @) Ex; ov B(nyp) we can wribe, ek) =) P trerd PlGix 8p [T-*) = Px,0 %)9-- 0% OTK] als PLt=k] = fot (2) pica = Cy Gime, es thy doesn’t depen wnbtin pw Ie doesn't degend enp | Téa, & soffident fy ts Thosiere Tfackorisatia n theme] “EE The necessary g sop fiden jep admit suPfiden statis tie provided by | theoem que Negman « statements i Te T(s) ie whfilent oP 9 ifs if aod only oF | tan be eaprened in the form eon diton fox a dibibuton a ckoris-ation int density Function Leg ire « hh) whee grea) Aepentt value of Te) ¢ \ independent: of & on & ¢ 2 only Ahrough theTIFT should be clenty @ \nleg: of mel Stee ean “hake ayo its demain rae a3 of con ain gy: rnin ete 9 SEEMS (2) the original Sample " alwag a sof fidenk for the prsameter woncesned ~ Invaviance property of soptident estimabry (fF T iv @ sop fid ent hina by o 4 Yr) is a one te one function ot Ts then > soffident fox ¥[e) . FAishexy— Neuman A oie sp the rgelyhood & if and only ges sample) be expieiied as 4 Le t F(a; 0) ie . see ® yh (%r- 40) e ae ave be tole sample fx %) Let Ky) For * -Xq oa Find ia ckimatoy M1 p4 : Cg Aten oe he peat Then, Lipabihood fanchlon » L = T ete) 135 te 0tae ee i= let (2 matenaie ee | exp C (7 ~ 2 R. pS * x) Here , t= =(E(), (0) = (44 ,e0 P he fry =, Here, Sx. R sophdenl Py pe qex % sufBuenk By Fromm normals popule bon tm Yip fy a sample of sxe o drawn , with mean p4 . : , Nayiqnce he ; ' — ierhene (Ms = t a Give an example fy himabvys which ave , iy ee anbised 4 eFfident i) @ arbiged 2 ineP iden F wy biased ¢ inePfidenly 1) S-7 Sample mean ig tof Rien Foy estima fing pol mean in passon distibution ) ot € it © %& Krown in random Sampling From nosmal diztributlon , % ie verlvent etimaby of fe > Bue, irk Krowpe, 52 =, 2 (4-3) 38 nol sup Aden oe fies. Suggest a suf fident es ima boy \ dial 6) Fax bemcull "with prob: oF sone Py the no. of Sacer ino tridle 1s supPidenl estima boy ae 2 §the vemml N(o,0")_ examine 8ffideny oP $4? “fy o in distribu Hen X ie wwf Adent eabioaty °F pp uhen samples ef dae ce) aye ~ bake from © BO 1p) ngwers + Let XX: y---Xq be the vandem sample Prem p(4) . F(t, 4) = ae MX, 20/1 ,23- Tr, Bigelihood Function, LOxy4) = Pre(a,4) , EA aye eve... Odes x) a! (G4%4+--+%) _n4 x ieasloaeaal 8) = 9 (t,x). be) = L(x) =a a : mr Tl --Apl nx -o4 = Ge eet [© 1 soffidené fer 4 8 Fy «a normal opula tion N(fse) 9 P(x 2 ree oD) tee ay Xn be sandom tam ple dwn From N(Fid Wey xX, X25 -- Then, Lifelihood Function is y L(y %a---%y fhe?) = Pls Hse) -P(tarpio’y... ve I Feisty) fal ey 7: x i=t 2 (aS “[oeay erie BES (qj +(e 922) Set, exp ees = 3 ~ P(tn fs)“ied see =e] ep (“2 os? ie fie Here, Punctlen io broken io wo pavts , Sewn one wotaing ¥ (cstimater) ¢ fb Cyarameter) » Fivel pave doer not contain R- | Tes, fox Rowing ca, ! &,) = TS sopfdent fr p- [ewevey , the fan facoy entries € s? anf ha % second part coobsin co? . Heme , 5? 1S noba tuphicienE * Jesbmaby Ry of. IF we ooo. = 4€ («iH Theo, a Funchlon wit be, Eee sorts) x Hence, S? Ty sop ficient ne Or 6. |q elton. Pie) = Pops ow, EGiay 2 Oe aiteres: yan} Ham f { ay NOD % F oro . Bt ei os 3 neR,o 30 ) ae Lipeldhe of feaction 5 (£0) = Tee 7) kd a 2. i eae lea ti = esr e eeStamey Kao _Tyequality szed eg Hnabor fox V9), FT ie ambi Function of 6% then N(t)4 fa ey7" ® "Ge S94] = Oey" (9) here Tie) ie dhe Fithey information on © , supplied by ample - 5 Semaxk? Cramer- Rao inequality ov ides 4 Qower bounk GNM 4> the variance of unbiased extimatny of Ve) . zi we make the followin cagorop Lion inorder pre Cramey— Kao inequality Lveguianity woditon: oy cramey—Rao inequality ; () he sovameter Bace 8 Www pon -degenenike open lnkewval on the ‘eal line - (2) For almost all (Xi -+ Xn) andl fpr all oc pubic drain GO ye SO bGre) exist , @ iE) “as of ifeysation 7% Wodepandent of} parmeten €. go that the f(x,e) ave Aipferenbiable unded ‘iokayral Sign - , (4) Be wodifons of aniforrn _convergene of the ‘ategre aye eatigfed to thar diffeventigtion unden integsaqn 1 vali | 5} 2(@) = EQ tay fas Ulve)) eve and sontive for dlg-eo le © be gondon vantable fororig te pat £ (190) pad [eeueter be te ae ee f de Sample a & the ae dhs gepulakion- Aun v2 1 G9) #f, £88) (i (149) dx Syl Le ws dhe {etut p dt fx] vas [ff ibaden ee soit. Aogert fp © ude de nequlauity Condittous pifercctiating lwo 9c (es 20 yi => Va bast) tage 0 © eye (2a) e ° @ Ld te bleotaa--28) \e an uabteed erttinator Ave) sul tat 00) = fib de - as(¢) _@ diatitag ne aotlh pepect 40. @ we ae [td de as(o) Ao 2s \eftear) Lde = ¥(0) > (ta4)-Ve)_ © We OO, —EEEaeeCah Ah) = a(t bgt) = 20) # =O) ote Fun O4@] we hau | to Zi ay (@& (ei) "2 vn@) “(ev Gs abal)) £ VO). ( At) > (vio) 4 VQ): (eabat)- - eusy)) Sve) 4 V(t). ©( Hash) E dkegt 0 e be re vi)> oD) eae Abe lool E (alegl Cty selec let 5A ay 5 ok Gdeguci f & 3 ee ak Fvleme A te Ange ve {(0¥0)- o= F(bs6) ae es de — $(ax6). ta + foe 62: APE Unktared oAtmale d, Ha pasameler © Wu €(1)-0 & v(6)-0 be Vle)= 1 2. vG)S i ne E(alaly! (0) du Ue) th the Aisle Caformacion alecut 8 supe nt sample: & 19)" E(A leat J: - &( Hel | Proc, * de Yung eek ts =H ‘ea a | Noo a mt bela) ) heat 2 4 | eg 8 (Shake 2 ee | ‘de 5 ) tt L\— dl J ee ecg iy | deOT 5 iF conditlons fox equality igo. ‘in Cramer Kao inequality * 4 Neve » v(t) J 6 Hoy . o 5 ee y @ [ey 2 ervey. (edge f — oy Te sign of equality ait) hold in ce Inequality 1F and o | . . : ‘one ‘E the vgn of equality holds in Tie ts hep pening enlg” YF the vaviabia (¢-V(e)) ¢ (2fogt aye 26 : Proportional > each obher &, Cr-viel = A[9) = 4 fay) Loe | Where A i dhe conibent} which T independent of 2%). x 2 Beg = (£39) = (b -Ve))- ajo (a Ie Alo =e where Ale) = a Hence, the neceseyy GF sophdent conditions foy unbiased egfj [bb obbtin the — lowey bound of its: variance Vy §iven by V(t) = Oley | & (2De Ly Qe But y (ig Ly = EfAle) (¢-v(0))]? = Gt)" €(4- voy" #8 taL$ = (Aey]* yn We gehy WH = Wort J (Aly VIE) — Nt) = | V0) A(e)Py ty CO) = (Ie). diy] — ee Hencey iF the kel hoo} nition L ix Satitfying @ » ther HD + Te vbiased atimboy of Ye) - (dD Minimum — vayience bound estimatwy [mvBE] exk ts - i) NCD) = [9M 4(0)| : Remark: ‘ OM Me)=e , a, F & & en, Unbiased estimatry of oy then @ xan be aritten a3), Mog L =b-e + They b® F de ee tee et YOO = [4 A(oy| 2 | Aye] | obkain VBE fox Rio Neyo) where go? & known at (FX Kay--y%q TS the eg of xe Sn’ From noymal [population » then likelihood Function i, n el Th F(x) 8) al . 0 : = - = GoW >= (Fe) eb ee ey | feeg t= _. leg (HI = LZ (ae HY" atte - K-32 eh 207 iz Pa 0 pt, Eto) Se ae tale x) BT x? = ieee ty) =o i eee (On) o* be : t=t Fae (i “) rot tea Ags, pees , “V(b) = Aei\ = Tal * ‘en bie. tale ; ae Fla2,0,1) “or eey, S TER “ — any TA e,9 ff : = eal ste wiae +e) ‘ n° Ag h = og ien) + 2 (ley tepees) = ples (Gay al lorry E> Z ley (4-04) 2 og “9 | at -s x 2q~- A cod ee i=y aoe Se) | = 2 2(r-9) | =| t+orey 7 : | a, Mer Ly Zee). 17 > 6 ET ae a0F oieie ince ib canncl eapressed 93 , Pag = (ee). . MB estimatoy doesn't exicl m & in pandas : So, Cfamey Fao lower bound jy sali b Ahe variance any anblagedF pA mS i002 % taken Prom Normal Fee Examine i 2% iJo ? i es mean 0 4 vaytance 2 - mvBE PY og? . 8 Leb %iom)..,%9, be the vs of fize O N(oys) - . : —(x) 2 Weje)= 1 ga —olLtL of r=e%% Ye)=o> aaa oe ‘taken From +e } = [on fofing logy Jog oe) =o fog Qasr *) —_1 hs (sree cee ia < x 26> p xen 1 gat i OL us = eee 20% 207 = -nis 2 —— — 3.43 a cay cane = Ste Qo" o- J 2 2 = Ue “yal a0 a - «(at ae j —— IF We of Foym b-V[0) -Ale) 207 N(H) = \V(0) gun (i = |i xaey 1 =e be = 20sie ae nites fle 2 ioe i "ve Eicie oP & & hax vaviance % { Leb %,95;.-%, be the ¥S a Prom ap: ges =e ~¥e Z Ygehe aan “Eye : °n Faking bog 5 Rog = 9 Qog fat ex, =—oley 9 - £2; MQogl OS Carat ¢ , e- sy St e 1 Seay Ss o eee e = Se; 4 Sr oh = (25 x) -0 x8 2 0 eo e a a ale Pi oF 1 12 VO) = | 90) -AG09| e #|Sladcw elliga CHamer Rao tnequality provider 4 techoique of Finding ie abo an MYO estimaloy ox ou tts application act the Same my an anbiared eatimetoy Here» vegalsvity wnditions aye very strich 9 peame quibe yestridive - moveovey y MVBE ix as NUE, sInCe the CR Lower band may ooh alvays be attained < “eveovey if the reqularity” conditions [oF trance ¥ Red] are vil@te}, then the Rew attaindle variane will be fey than CR bound + Heve, we bry bo obtgig MNu esfimabey From ary anti akimateys AhYough the vse of suffident statistic: Wis rechnique Is called blackwellisation |Rao- Blackwell Theoyem f Leb x&y be v's och Hab EV)= € NY) =y Go) Let E(Vfx22) = $(x) > then 0) (pe) = pp i) V(g@) 2 v(Y) Tik i the Rao- Blackwell theovem-