On Estimation of Unknown Parameters of Exponential- Logarithmic Distribution by Censored Data

Similar documents
Reliability of time dependent stress-strength system for various distributions

Suzan Mahmoud Mohammed Faculty of science, Helwan University

Unbalanced Panel Data Models

3.4 Properties of the Stress Tensor

Consistency of the Maximum Likelihood Estimator in Logistic Regression Model: A Different Approach

MODEL QUESTION. Statistics (Theory) (New Syllabus) dx OR, If M is the mode of a discrete probability distribution with mass function f

LECTURE 6 TRANSFORMATION OF RANDOM VARIABLES

Lecture 1: Empirical economic relations

Introduction to logistic regression

Estimation of the Present Values of Life Annuities for the Different Actuarial Models

COMPLEX NUMBERS AND ELEMENTARY FUNCTIONS OF COMPLEX VARIABLES

Bayesian Shrinkage Estimator for the Scale Parameter of Exponential Distribution under Improper Prior Distribution

Binary Choice. Multiple Choice. LPM logit logistic regresion probit. Multinomial Logit

Tolerance Interval for Exponentiated Exponential Distribution Based on Grouped Data

Odd Generalized Exponential Flexible Weibull Extension Distribution

ASYMPTOTIC AND TOLERANCE 2D-MODELLING IN ELASTODYNAMICS OF CERTAIN THIN-WALLED STRUCTURES

Department of Mathematics and Statistics Indian Institute of Technology Kanpur MSO202A/MSO202 Assignment 3 Solutions Introduction To Complex Analysis

Numerical Method: Finite difference scheme

Counting the compositions of a positive integer n using Generating Functions Start with, 1. x = 3 ), the number of compositions of 4.

Second Handout: The Measurement of Income Inequality: Basic Concepts

Estimation Theory. Chapter 4

A Stochastic Approximation Iterative Least Squares Estimation Procedure

Repeated Trials: We perform both experiments. Our space now is: Hence: We now can define a Cartesian Product Space.

Total Prime Graph. Abstract: We introduce a new type of labeling known as Total Prime Labeling. Graphs which admit a Total Prime labeling are

Chapter 6. pn-junction diode: I-V characteristics

The real E-k diagram of Si is more complicated (indirect semiconductor). The bottom of E C and top of E V appear for different values of k.

Bayesian Test for Lifetime Performance Index of Ailamujia Distribution Under Squared Error Loss Function

Three-Dimensional Theory of Nonlinear-Elastic. Bodies Stability under Finite Deformations

Almost all Cayley Graphs Are Hamiltonian

Aotomorphic Functions And Fermat s Last Theorem(4)

Chapter 5 Special Discrete Distributions. Wen-Guey Tzeng Computer Science Department National Chiao University

Math Tricks. Basic Probability. x k. (Combination - number of ways to group r of n objects, order not important) (a is constant, 0 < r < 1)

Correlation in tree The (ferromagnetic) Ising model

β-spline Estimation in a Semiparametric Regression Model with Nonlinear Time Series Errors

Chiang Mai J. Sci. 2014; 41(2) 457 ( 2) ( ) ( ) forms a simply periodic Proof. Let q be a positive integer. Since

Introduction to logistic regression

The R Package PK for Basic Pharmacokinetics

A COMPARISON OF SEVERAL TESTS FOR EQUALITY OF COEFFICIENTS IN QUADRATIC REGRESSION MODELS UNDER HETEROSCEDASTICITY

Ordinary Least Squares at advanced level

ERDOS-SMARANDACHE NUMBERS. Sabin Tabirca* Tatiana Tabirca**

Different types of Domination in Intuitionistic Fuzzy Graph

Complex Numbers. Prepared by: Prof. Sunil Department of Mathematics NIT Hamirpur (HP)

In 1991 Fermat s Last Theorem Has Been Proved

Review - Probabilistic Classification

On the Possible Coding Principles of DNA & I Ching

The probability of Riemann's hypothesis being true is. equal to 1. Yuyang Zhu 1

Channel Capacity Course - Information Theory - Tetsuo Asano and Tad matsumoto {t-asano,

BER Analysis of Optical Wireless Signals through Lognormal Fading Channels with Perfect CSI

A Measure of Inaccuracy between Two Fuzzy Sets

Independent Domination in Line Graphs

Estimation of Population Variance Using a Generalized Double Sampling Estimator

MODELING TRIVARIATE CORRELATED BINARY DATA

The Beta Inverted Exponential Distribution: Properties and Applications

On the Beta Mekaham Distribution and Its Applications. Chukwu A. U., Ogunde A. A. *

The Hyperelastic material is examined in this section.

Optimal Progressive Group-Censoring Plans for. Weibull Distribution in Presence. of Cost Constraint

X ε ) = 0, or equivalently, lim

ME 501A Seminar in Engineering Analysis Page 1

Jurnal Teknologi HYPOTHESIS TESTING FOR THE PARAMETERS OF LOG-LOGISTIC REGRESSION MODEL WITH LEFT- TRUNCATED AND RIGHT-CENSORED SURVIVAL DATA

Machine Learning. Principle Component Analysis. Prof. Dr. Volker Sperschneider

Irregular Boundary Area Computation. by Quantic Hermite Polynomial

Weights Interpreting W and lnw What is β? Some Endnotes = n!ω if we neglect the zero point energy then ( )

i j i i i = S i 1 Y Y i F i ..., X in

Comparing Different Estimators of three Parameters for Transmuted Weibull Distribution

1985 AP Calculus BC: Section I

Dr. Junchao Xia Center of Biophysics and Computational Biology. Fall /21/2016 1/23

Chapter 5 Properties of a Random Sample

International Journal of Advanced Scientific Research and Management, Volume 3 Issue 11, Nov

Lecture #11. A Note of Caution

Parameter, Statistic and Random Samples

Phase diagram and frustration of decoherence in Y-shaped Josephson junction networks. D.Giuliano(Cosenza), P. Sodano(Perugia)


signal amplification; design of digital logic; memory circuits

Chapter 4 NUMERICAL METHODS FOR SOLVING BOUNDARY-VALUE PROBLEMS

Folding of Regular CW-Complexes

FOURIER SERIES. Series expansions are a ubiquitous tool of science and engineering. The kinds of

Chapter 14 Logistic Regression Models

CHAPTER VI Statistical Analysis of Experimental Data

MATHEMATICAL IDEAS AND NOTIONS OF QUANTUM FIELD THEORY. e S(A)/ da, h N

Graphs of q-exponentials and q-trigonometric functions

On Approximation Lower Bounds for TSP with Bounded Metrics

Note on the Computation of Sample Size for Ratio Sampling

STK4011 and STK9011 Autumn 2016

Nuclear Chemistry -- ANSWERS

Point Estimation: definition of estimators

Power Spectrum Estimation of Stochastic Stationary Signals

Entropy Equation for a Control Volume

More Statistics tutorial at 1. Introduction to mathematical Statistics

Section 11.6: Directional Derivatives and the Gradient Vector

Linear Algebra Existence of the determinant. Expansion according to a row.

CHAPTER: 3 INVERSE EXPONENTIAL DISTRIBUTION: DIFFERENT METHOD OF ESTIMATIONS

Integral points on hyperbolas over Z: A special case

Chapter 6 Student Lecture Notes 6-1

Almost unbiased exponential estimator for the finite population mean

Law of large numbers

BAYESIAN ANALYSIS OF THE SIMPLE LINEAR REGRESSION WITH MEASUREMENT ERRORS

Construction of asymmetric orthogonal arrays of strength three via a replacement method

On Three-Way Unbalance Nested Analysis of Variance

APPENDIX: STATISTICAL TOOLS

Continuous Distributions

Transcription:

saqartvlos mcrbata rovul akadms moamb, t 9, #2, 2015 BULLETIN OF THE GEORGIAN NATIONAL ACADEMY OF SCIENCES, vol 9, o 2, 2015 Mathmatcs O Estmato of Ukow Paramtrs of Epotal- Logarthmc Dstrbuto by Csord Data Al Pjya Dpartmt of Mathmatcs, Faculty of Eact ad Natural Sccs, Tbls Stat Uvrsty, Tbls (Prstd by Acadmy Mmbr Elzbar Nadaraya) ABSTRACT Th problm of stmato of paramtrs of Epotal-Logarthmc dstrbuto th cas of csord data s cosdrd W usd psudo mamum lklhood mthod ad costructd a procdur to solv ths problm Thorm of cosstcy s provd Smulato s usd to study th proprts of stmators drvd 2015 Bull Gorg Natl Acad Sc Ky words: Epotal-Logarthmc dstrbuto, psudo mamum lklhood stmators, cosstt stmators, partly csord data Th Epotal Logarthmc (EL) dstrbuto s a famly of lftm dstrbutos wth dcrasg falur rat, dfd o th trval [0, ), s [1] Ths dstrbuto s paramtrzd by two paramtrs: p (0,1) ad >0 (EL) dstrbuto s usd th study of lgths of orgasms, dvcs, matrals, tc, whch s of major mportac th bologcal ad grg sccs Th gv work studs th problm of paramtrs stmato of (EL) dstrbuto wth a spcfc obsrvato pattr Th obsrvatos ar assumd to group ad full o obsrvatos of dvdual ralzato valus ar possbl Applcato of a modfd a crta ss, mthod of mamum lklhood s suggstd ad th offrd procdur s show to lad to cosstt stmators Lt X b a radom varabl wth a dstrbuto fucto F( ) F(, ), whr s a ukow vctor paramtr a ft dmsoal Euclda spac q R Suppos s compact W hav to costruct a cosstt stmator basd o obsrvatos of a radom varabl X Th prmt s dsgd such a way that th actual umbr of ralzatos s ukow to us, w kow oly a part of thos ralzatos Lt fd pots 0 t1 t2 t b gv o th l R (cludg th cas wh th last pot taks a ft valu) Ths pots mak thr catgors of trvals: 0) Th trval ( t, t 1) blogs to zro catgory f w do ot kow thr th dvdual valus of th sampl, or th umbr of sampl valus of th radom varabl apparg ths trval 1) Th trval ( t, t 1) blogs to th frst catgory f dvdual sampl valus ar ukow, but w 2015 Bull Gorg Natl Acad Sc

34 Al Pjya kow th umbr of sampl valus of th radom varabl X ths trval As usual, w dot ths umbr by 2) Th trval ( t, t 1) blogs to th scod catgory f w kow dvdual sampl valu 1, 2,, Furthr w dot summato ad tgrato th trvals of th zro, frst ad scod catgors by,(1) ad (2), rspctvly (s [2]) W call ths typ of sampl partally groupd sampl wth csorg Evdtly, csord sampls of both typs as wll as trucatd sampls mak a spcal cas of th statd problm Absc of formato th zro catgory trval crats crta dffcults, whch w try to ovrcom basd o our kowldg of th dstrbuto typ of F(, ) ad th umbr of sampl mmbrs that do ot appar th zro catgory trval: (1)(2) Lt A ( t, t 1) b zro catgory trval W dot by m th umbr of sampl mmbrs apparg A Th, r m th total umbr of obsrvatos Not that m m apparg A If F ( ) r F r (, ) dots a mprcal dstrbuto fucto, th m F ( r t 1) F r ( t ) m s, rlatv frqucy of X ad by vrtu of Broull s law of larg umbrs t covrgs to p ( ) F( t 1, ) F( t, ) wth probablty 1 By summato of Equalts (1) ovr all zro catgory trvals w obta whch lads to F r ( t 1) F r ( t ) m, 1 F ( r t 1) F r ( t ) F r ( t 1) F r ( t ) m (2) 1 ( F r t 1) F r ( t ) W td to apply th mthod of psudo mamum lklhood Assum that th radom varabl X has a dstrbuto dsty wth rspct to Lbsgu masur f ( ) f (, ) Th th lklhood fucto has th followg form: j (1) j1 m 1 1, (3) L ( ; ) F( t ) F(t ) F( t ) F(t ) f ( ) whr m, ar dfd by Formulas (2) Fdg mamum pots of th fucto L( ; ) bcoms complcatd sc t s dffcult to study smoothss proprts of mprcal fuctos Thrfor, w cosdr a modfd lklhood fucto: F ( t ) F ( t ) 1 1 F ( t1 ) F ( t ) 1 1 (4) L ( ; ) F( t ) F(t ) F( t ) F(t ) f ( ) j (1) j1 (1) Bull Gorg Natl Acad Sc, vol 9, o 2, 2015

O Estmato of Ukow Paramtrs of Epotal-Logarthmc Dstrbuto 35 Lmma: Lt th followg codtos b satsfd: a) Th dstrbuto fucto F(, ) s cotuous both varabl ad has a cotuous drvatv F (, ) f (, ) ; b) Th fucto L ( ; ) has a absolut mamum Th s cosstt ad asymptotcally ffct stmator of th tru valu of th paramtr 0 Th proof rsults from th corrspodg thorms of [3 4] Estmato of th paramtrs p ad of (EL) Dstrbuto Lt X b a radom varabl dstrbutd wth rspct to (EL) dstrbuto, wth dsty ad dstrbuto fucto 1 (1 p) f (, p, ) l p 1 (1 p) l(1 (1 p) ) F( ) 1, l p p (0,1), 0, [0, ) Lt [1, ) b a zro catgory trval Howvr, w hav obsrvatos X1, X 2,, X th trval [0,1) W hav to stmat ad p paramtrs wth rspct to ths obsrvatos W apply psudo mamum lklhood stmats I ordr to costruct lklhood fuctos ot that f w dot by k th umbr of gral sampl mmbrs that appar [1, ), th k wll b th frqucy k of thm apparg th trval [1, ) Thrfor by Broull Kolmogorov thorm w hav F( ) F(1) k 1 F( ) F(1) Sc th probablty of k lmts of th sampl apparg th black hol s [F( ) F(1)] k, w ca wrt th psudo lklhood fucto as Hc, l(1 (1 p) ) l pl(1 (1 p) ) l(1 (1 p) ) 1 (1 p) L ( ) (5) l p 1 l p 1 (1 p) l(1 (1 p) ) l(1 (1 p) ) 1 l L l l( ) l(1 p) l p l(1 (1 p) ) l p l p l l (1 (1 p) ) 1 1 Computg l L w gt Bull Gorg Natl Acad Sc, vol 9, o 2, 2015

36 Al Pjya (1 p) (1 p) (1 (1 p) )(l p l(1 (1 p) )) 1 (1 p) 1 1 (1 p) l p l(1 (1 p) ) l 2 (6) 1 (1 p) )(l p l(1 (1 p) )) l p l L Smlarly, p l(1 (1 p) ) l l p p p p p p 2 (l l(1 (1 ) )) (l l(1 (1 ) ))l Studyg (6), wh 0 ad w gt l p l(1 (1 p) ) 1 (1 p) p (7) p l p 1 p 1 (1 p) 1 l L lm 0 ad l L lm 1 Th cotuous fucto l L chags ts sg ad hc thr sts a pot, such that l L ( ) 0 W com to th sam cocluso rlatd to paramtr p: l L lm p 0 p ad l L lm p p 2 1 Th cotuous fucto l L p chags ts sg ad hc thr sts a pot p, such that l L ( p p ) 0 p O th bass of abov coclusos ad takg to cosdrato Lmma w ca stat that th followg thorms of cosstcy ar tru: Thorm 1: Lt X1, X 2,, X b radom varabls of a sampl dstrbutd wth rspct to (EL) dstrbuto wth ukow paramtr ad kow p Th obsrvatos ar mad o th trval [0,1) ad [1, ) thr sampl mmbrs, or thr umbr ar rcordd Th th psudo mamum lklhood stmator for paramtr sts ad s a uqu root of th quato: Bull Gorg Natl Acad Sc, vol 9, o 2, 2015

O Estmato of Ukow Paramtrs of Epotal-Logarthmc Dstrbuto 37 ad th stmator s cosstt (1 p) (1 p) (1 (1 p) )(l p l(1 (1 p) )) 1 (1 p) 1 1 (1 p) l p l(1 (1 p) ) l 0, 2 1 (1 p) )(l p l(1 (1 p) )) l p Thorm 2: Lt X1, X 2,, X b radom varabls of a sampl dstrbutd wth rspct to (EL) dstrbuto wth ukow paramtr p ad kow Th obsrvatos ar mad o th trval [0,1) ad [1, ) thr sampl mmbrs or thr umbr ar rcordd Th th psudo mamum lklhood stmator for p paramtr sts ad s a uqu root of th quato: l(1 (1 p) ) l l p p p p p p 2 (l l(1 (1 ) )) (l l(1 (1 ) ))l l p l(1 (1 p) ) 1 (1 p) p ad th stmator s cosstt 0, p l p 1 p 1 (1 p) 1 Tabl Th rsult of th prmt β p =100 =250 =500 =1000 p=02 458168887 444823007 435079164 42762195 p=05 438954193 422439165 417349481 40995178 p=07 453690541 439316409 419231157 40816133 β=5 172190903 248308392 131192655 08477539 β=6 652284317 120087317 295259175 05622895 β=7 225707934 473803005 138997548 37100971 Smulato Th objctv of our smulato s to chck rsults rcvd prcdg scto ad to compar th actual paramtrs valus wth valus, whch wr gott va prmts I ths study, th sampl sz chos s = 100, 250, 500 ad 1000 W cosdr svral cass for both ad p paramtrs Estmato rsults ar show th tabl blow Frst colum shows stmatd paramtrs ad th scod colum w cosdr svral cass for paramtr stmatg Th rsults show th Tabl mply that usg ths mthod stmators hav thr rror dcrasg as th sampl sz crass Bull Gorg Natl Acad Sc, vol 9, o 2, 2015

38 Al Pjya matmatka qspocalur-logartmul gaawlbs ucob paramtrbs Sfasba czurrbul moacmbt a pja javasvls salobs Tblss salmwfo uvrstts zust da sabubsmtyvlo mcrbata fakultts matmatks mmartulba (warmodgla akadms wvrs adaraas mr) statas warmodgla qspocalur-logartmul gaawlbs ucob paramtrbs Sfasba czurrbul moacmbt REFERENCES 1 Tahmasb R, Rza S, (2008) Computatoal Statstcs ad Data Aalyss, 52(8): 3889 3901 2 Kulldorff G, Cotrbutos to th thory of stmato from groupd ad partally groupd sampls ALMQVIST&WIKSELL, STOCKHOLM, GOTEBORG UPPSALA 3 Frguso TS (1996) A Cours larg sampl thory Chapma&Hall 4 Dllo JV, Lbao G (2010) Statstcal ad Computatoal Tradoffs Stochastc Compos Lklhood arxv: 10030691,v1:29 p Rcvd July, 2015 Bull Gorg Natl Acad Sc, vol 9, o 2, 2015