Tolerance Limits on Order Statistics in Future Samples Coming from the Two-Parameter Exponential Distribution

Similar documents
Chapter 6 Sampling Distributions

MOST PEOPLE WOULD RATHER LIVE WITH A PROBLEM THEY CAN'T SOLVE, THAN ACCEPT A SOLUTION THEY CAN'T UNDERSTAND.

FACULTY OF MATHEMATICAL STUDIES MATHEMATICS FOR PART I ENGINEERING. Lectures

Statistics 511 Additional Materials

5. Likelihood Ratio Tests

Confidence interval for the two-parameter exponentiated Gumbel distribution based on record values

Last Lecture. Wald Test

Bayesian Control Charts for the Two-parameter Exponential Distribution

A statistical method to determine sample size to estimate characteristic value of soil parameters

Lecture 6 Simple alternatives and the Neyman-Pearson lemma

2 1. The r.s., of size n2, from population 2 will be. 2 and 2. 2) The two populations are independent. This implies that all of the n1 n2

Resampling Methods. X (1/2), i.e., Pr (X i m) = 1/2. We order the data: X (1) X (2) X (n). Define the sample median: ( n.

MATH 320: Probability and Statistics 9. Estimation and Testing of Parameters. Readings: Pruim, Chapter 4

Statistical Intervals for a Single Sample

EXAMINATIONS OF THE ROYAL STATISTICAL SOCIETY

Estimation for Complete Data

Lecture 7: Properties of Random Samples

GUIDELINES ON REPRESENTATIVE SAMPLING

If, for instance, we were required to test whether the population mean μ could be equal to a certain value μ

Double Stage Shrinkage Estimator of Two Parameters. Generalized Exponential Distribution

Bootstrap Intervals of the Parameters of Lognormal Distribution Using Power Rule Model and Accelerated Life Tests

1036: Probability & Statistics

R. van Zyl 1, A.J. van der Merwe 2. Quintiles International, University of the Free State

Statistical Inference (Chapter 10) Statistical inference = learn about a population based on the information provided by a sample.

Estimation of a population proportion March 23,

BIOS 4110: Introduction to Biostatistics. Breheny. Lab #9

PREDICTION INTERVALS FOR FUTURE SAMPLE MEAN FROM INVERSE GAUSSIAN DISTRIBUTION

Topic 9: Sampling Distributions of Estimators

Chapter 6 Principles of Data Reduction

Topic 9: Sampling Distributions of Estimators

PSYCHOLOGICAL RESEARCH (PYC 304-C) Lecture 9

Stat 319 Theory of Statistics (2) Exercises

Statistical inference: example 1. Inferential Statistics

A NEW METHOD FOR CONSTRUCTING APPROXIMATE CONFIDENCE INTERVALS FOR M-ESTU1ATES. Dennis D. Boos

Discrete Mathematics for CS Spring 2008 David Wagner Note 22

- E < p. ˆ p q ˆ E = q ˆ = 1 - p ˆ = sample proportion of x failures in a sample size of n. where. x n sample proportion. population proportion

Extreme Value Charts and Analysis of Means (ANOM) Based on the Log Logistic Distribution

Frequentist Inference

1.010 Uncertainty in Engineering Fall 2008

Stat 421-SP2012 Interval Estimation Section

6 Sample Size Calculations

Investigating the Significance of a Correlation Coefficient using Jackknife Estimates

Probability and statistics: basic terms

Random Variables, Sampling and Estimation

Some Properties of the Exact and Score Methods for Binomial Proportion and Sample Size Calculation

Chapter 8: STATISTICAL INTERVALS FOR A SINGLE SAMPLE. Part 3: Summary of CI for µ Confidence Interval for a Population Proportion p

7-1. Chapter 4. Part I. Sampling Distributions and Confidence Intervals

Topic 9: Sampling Distributions of Estimators

Tests of Hypotheses Based on a Single Sample (Devore Chapter Eight)

NCSS Statistical Software. Tolerance Intervals

Read through these prior to coming to the test and follow them when you take your test.

Direction: This test is worth 150 points. You are required to complete this test within 55 minutes.

Access to the published version may require journal subscription. Published with permission from: Elsevier.

Sample Size Determination (Two or More Samples)

Continuous Data that can take on any real number (time/length) based on sample data. Categorical data can only be named or categorised

Chapter 22. Comparing Two Proportions. Copyright 2010, 2007, 2004 Pearson Education, Inc.

A goodness-of-fit test based on the empirical characteristic function and a comparison of tests for normality

On an Application of Bayesian Estimation

Econ 325 Notes on Point Estimator and Confidence Interval 1 By Hiro Kasahara

This chapter focuses on two experimental designs that are crucial to comparative studies: (1) independent samples and (2) matched pair samples.

April 18, 2017 CONFIDENCE INTERVALS AND HYPOTHESIS TESTING, UNDERGRADUATE MATH 526 STYLE

This is an introductory course in Analysis of Variance and Design of Experiments.

Agreement of CI and HT. Lecture 13 - Tests of Proportions. Example - Waiting Times

Properties and Hypothesis Testing

Goodness-Of-Fit For The Generalized Exponential Distribution. Abstract

GG313 GEOLOGICAL DATA ANALYSIS

Comparing Two Populations. Topic 15 - Two Sample Inference I. Comparing Two Means. Comparing Two Pop Means. Background Reading

Overview. p 2. Chapter 9. Pooled Estimate of. q = 1 p. Notation for Two Proportions. Inferences about Two Proportions. Assumptions

Simulation. Two Rule For Inverting A Distribution Function

Statisticians use the word population to refer the total number of (potential) observations under consideration

Confidence Interval for Standard Deviation of Normal Distribution with Known Coefficients of Variation

STAT431 Review. X = n. n )

Chapter 22. Comparing Two Proportions. Copyright 2010 Pearson Education, Inc.

Math 140 Introductory Statistics

MOMENT-METHOD ESTIMATION BASED ON CENSORED SAMPLE

Mathematical Modeling of Optimum 3 Step Stress Accelerated Life Testing for Generalized Pareto Distribution

( θ. sup θ Θ f X (x θ) = L. sup Pr (Λ (X) < c) = α. x : Λ (x) = sup θ H 0. sup θ Θ f X (x θ) = ) < c. NH : θ 1 = θ 2 against AH : θ 1 θ 2

Thomas Whitham Form Centre

Maximum likelihood estimation from record-breaking data for the generalized Pareto distribution

EFFECTIVE WLLN, SLLN, AND CLT IN STATISTICAL MODELS

(7 One- and Two-Sample Estimation Problem )

1 Inferential Methods for Correlation and Regression Analysis

Expectation and Variance of a random variable

HYPOTHESIS TESTS FOR ONE POPULATION MEAN WORKSHEET MTH 1210, FALL 2018

Common Large/Small Sample Tests 1/55

Sample Size Estimation in the Proportional Hazards Model for K-sample or Regression Settings Scott S. Emerson, M.D., Ph.D.

CONTROL CHARTS FOR THE LOGNORMAL DISTRIBUTION

Mathacle. PSet Stats, Concepts In Statistics Level Number Name: Date: Confidence Interval Guesswork with Confidence

Basis for simulation techniques

STAC51: Categorical data Analysis

Department of Mathematics

Introduction to Econometrics (3 rd Updated Edition) Solutions to Odd- Numbered End- of- Chapter Exercises: Chapter 3

Section 9.2. Tests About a Population Proportion 12/17/2014. Carrying Out a Significance Test H A N T. Parameters & Hypothesis

KLMED8004 Medical statistics. Part I, autumn Estimation. We have previously learned: Population and sample. New questions

Bayesian and E- Bayesian Method of Estimation of Parameter of Rayleigh Distribution- A Bayesian Approach under Linex Loss Function

DS 100: Principles and Techniques of Data Science Date: April 13, Discussion #10

Lecture 18: Sampling distributions

Class 23. Daniel B. Rowe, Ph.D. Department of Mathematics, Statistics, and Computer Science. Marquette University MATH 1700

STA6938-Logistic Regression Model

Exam II Review. CEE 3710 November 15, /16/2017. EXAM II Friday, November 17, in class. Open book and open notes.

Transcription:

America Joural of Theoretical ad Applied tatistics 6; 5(-): -6 Published olie November 8 5 (http://wwwsciecepublishiggroupcom/j/ajtas) doi: 648/jajtass65 IN: 36-8999 (Prit); IN: 36-96 (Olie) Tolerace imits o Order tatistics i Future amples Comig from the Two-Parameter Expoetial Distributio Nicholas A Nechval * Kostati N Nechval Departmet of Mathematics Baltic Iteratioal Academy Riga atvia Departmet of Applied Mathematics Trasport ad Telecommuicatio Istitute Riga atvia Email address: echval@juilv (N A Nechval) osta@tsilv (K N Nechval) To cite this article: Nicholas A Nechval Kostati N Nechval Tolerace imits o Order tatistics i Future amples Comig from the Two-Parameter Expoetial Distributio America Joural of Theoretical ad Applied tatistics pecial Issue: Novel Ideas for Efficiet Optimizatio of tatistical Decisios ad Predictive Ifereces uder Parametric certaity of derlyig Models with Applicatios Vol 5 No - 6 pp -6 doi: 648/jajtass65 Abstract: This paper presets a iovative approach to costructig lower ad upper tolerace limits o order statistics i future samples Attetio is restricted to ivariat families of distributios uder parametric ucertaity The approach used here emphasizes pivotal quatities relevat for obtaiig tolerace factors ad is applicable wheever the statistical problem is ivariat uder a group of trasformatios that acts trasitively o the parameter space It does ot require the costructio of ay tables ad is applicable whether the past data are complete or Type II cesored The proposed approach requires a quatile of the F distributio ad is coceptually simple ad easy to use For illustratio the two-parameter expoetial distributio is cosidered A practical example is give Keywords: Order tatistics F Distributio ower Tolerace imit pper Tolerace imit Itroductio tatistical tolerace limits are aother tool for maig statistical iferece o a uow populatio As opposed to a cofidece limit that provides iformatio cocerig a uow populatio parameter a tolerace limit provides iformatio o the etire populatio; to be specific oe-sided tolerace limit is expected to capture a certai proportio or more of the populatio with a give cofidece level For example a upper tolerace limit for a uivariate populatio is such that with a give cofidece level a specified proportio or more of the populatio will fall below the limit A lower tolerace limit satisfies similar coditios It is ofte desirable to have statistical tolerace limits available for the distributios used to describe time-to-failure data i reliability problems For example oe might wish to ow if at least a certai proportio say of a maufactured product will operate at least T hours This questio ca ot usually be aswered exactly but it may be possible to determie a lower tolerace limit ( ) based o a prelimiary radom sample ( ) such that oe ca say with a certai cofidece that at least % of the product will operate loger tha ( ) The reliability statemets ca be made based o ( ) or decisios ca be reached by comparig ( ) to T Tolerace limits of the type metioed above are cosidered i this paper That is if f (x) deotes the desity fuctio of the paret populatio uder cosideratio ad if is ay statistic obtaied from the prelimiary radom sample ( ) of that populatio the () is a lower probability tolerace limit for proportio if Pr f ( xdx ) () ( ) ad () is a upper probability tolerace limit for proportio if ( ) Pr f ( xdx ) () is the parameter (i geeral vector) The commo distributios used i life testig problems are the ormal expoetial Weibull ad gamma distributios [] Tolerace limits for the ormal distributio have bee

Nicholas A Nechval ad Kostati N Nechval: Tolerace imits o Order tatistics i Future amples Comig from the Two-Parameter Expoetial Distributio cosidered i [] [3] [4] ad others Tolerace limits ejoy a fairly rich history i the literature ad have a very importat role i egieerig ad maufacturig applicatios Patel [5] provides a review (which was fairly comprehesive at the time of publicatio) of tolerace itervals for may distributios as well as a discussio of their relatio with cofidece itervals for percetiles ad predictio itervals Dusmore [6] ad Guether Patil ad ppuluri [7] both discuss -parameter expoetial tolerace itervals ad the estimatio procedure i greater detail Egelhardt ad Bai [8] discuss how to modify the formulas whe dealig with type II cesored data Guether [9] ad Hah ad Meeer [] discuss how oe-sided tolerace limits ca be used to obtai approximate two-sided tolerace itervals by applyig Boferroi's iequality I cotrast to other statistical limits commoly used for statistical iferece the tolerace limits (especially for the order statistics) are used relatively rarely Oe reaso is that the theoretical cocept ad computatioal complexity of the tolerace limits is sigificatly more difficult tha that of the stadard cofidece ad predictio limits Thus it becomes ecessary to use the iovative approaches which will allow oe to costruct tolerace limits o future order statistics for may populatios I this paper the iovative approach to costructig lower ad upper tolerace limits o order statistics i future samples is proposed For illustratio the two-parameter expoetial distributio is cosidered Mathematical Prelimiaries Probability Distributio Fuctio of Order tatistic Theorem If there is a radom sample of m ordered observatios Y Y m from a ow distributio (cotiuous or discrete) with desity fuctio f (y) distributio fuctio F (y) the the probability distributio fuctio of the th order statistic Y { m} is give by m m j PY ( y m) [ F ( y )] [ F ( y )] j j mj f ( xdx ) ( m + ) (3) F ( y ) F ( y ) ( m + ) [( m + ) + ]/ ( m + ) x + x > is the probability desity fuctio of a F distributio with (m+) ad degrees of freedom Proof uppose a evet occurs with probability p per trial It is well-ow that the probability P of its occurrig or more times i m trials is termed a cumulative biomial probability ad is related to the icomplete beta fuctio I x (a b) as follows: (4) m m j mj P p ( p) I ( m + ) p j (5) j It follows from (5) that m m j P { Y y m} [ F ( y )] [ F ( y )] j j I ( m + ) ( ) F y F( y ) ( m + ) u ( u) du Β( m + ) ( m + )/ ( m ) + ( m + ) Β ( m + )/ F ( y ) ( m + ) + u mj u du u ( m + ) ( m + ) u ( m + )/ ( m ) + ( m + ) Β F ( y ) F ( y) ( m + ) x [( m + ) + ]/ ( m + ) x + dx ( m + )/ (6) f ( x) + ( m + ) Β ( m ) This eds the proof Corollary u x u ( m + ) (7) ( m + ) ( m + ) x ( m + )/ m j PY ( > y m) [ F ( y )] [ F ( y )] j j mj

America Joural of Theoretical ad Applied tatistics 6; 5(-): -6 3 F ( y ) F ( y ) ( m + ) f ( m + ) ( xdx ) Corollary If y m; is the quatile of order for the distributio of Y we have from (3) that y m; is the solutio of ( m + ) ; (8) F ( y ) m ; (9) + ( m + ) q ; is the quatile of order for the F distributio with (m+) ad degrees of freedom Two-Parameter Expoetial Distributio The two-parameter expoetial distributio is a widely used ad widely ow distributio It is characterized by the desity fuctio is the pivotal quatity with the desity fuctio gw w w Γ( r ) r ( ) exp( ) w 3 Tolerace imits for Order tatistic 3 over Tolerace imit (8) Theorem et r be the first r ordered observatios from the prelimiary sample of size from a two-parameter expoetial distributio defied by the desity fuctio () The a lower oe-sided -cotet tolerace limit at level () (o the th order statistic Y from a set of m future ordered observatios Y Y m also from the distributio () ) which satisfies ( PY m ) Pr ( > ) (9) yµ f ( y) exp y µ σ σ () (µσ) σ (scale parameter) ad µ (shift parameter) are uow The distributio fuctio of the two-parameter expoetial distributio is yµ F ( y) exp σ () The sufficiet statistic for the parameter based o the r ( ) smallest observatios ( r ) i a radom sample of size from the two-parameter expoetial distributio () is r mi( ) + ( r ) () r i r i x ( µ ) ~ h ( x ) exp x µ σ σ V (3) µ (4) σ is the pivotal quatity with the desity fuctio ( ) hv ( ) exp v v (5) s r ~ g ( s ) s exp s σ r Γ( r ) σ σ (6) is give by r l( ) + if l r l( ) if < l ( m + ) q( m + ) ; ( m + ) q + ( m + ) ; Proof It follows from (8) () ad (9) that ( PY > m ) Pr ( ) F ( ) F ( ) ( m + ) Pr f ( xdx ) ( m + ) F ( ) Pr q ( m + ) ; F ( ) ( m + ) Pr F ( ) ( m ) q + + ( m + ) ; () () W (7) σ µ ( m ) q + ( m + ) ; Pr exp σ ( m ) q + + ( m + ) ;

4 Nicholas A Nechval ad Kostati N Nechval: Tolerace imits o Order tatistics i Future amples Comig from the Two-Parameter Expoetial Distributio µ ( m ) q + m Pr l ( ) ( + ) ; σ m q + + ( m + ) ; ( l ) PrV W l w hvdv ( ) () r l + if l r l if < l (8) is the lower tolerace factor (3) ( m + ) q( m + ) ; ( m + ) q + ( m + ) ; (9) ( m + ) q( m + ) ; ( m + ) q + ( m + ) ; It follows from (8) (9) ad () that l w arg hvgwdvdw ( ) ( ) (4) Proof It follows from (3) () ad (7) that ( PY m ) Pr ( ) F ( ) F( ) ( m + ) Pr f ( xdx ) ( m + ) /( r) (5) Taig ito accout (3) ad (5) we have () This completes the proof Corollary If m the r l( ) + if l r l( ) if < l The result similar to that of (6) ca be foud i [7 8] 3 pper Tolerace imit (6) Theorem 3 et r be the first r ordered observatios from the prelimiary sample of size from a two-parameter expoetial distributio defied by the desity fuctio () The a lower oe-sided -cotet tolerace limit at level () (o the th order statistic Y from a set of m future ordered observatios Y Y m also from the distributio () ) which satisfies is give by ( PY m ) Pr ( ) (7) F ( ) F ( ) ( m + ) Pr f ( xdx ) ( m + ) F ( ) Pr q ( m + ) ; F ( ) ( m + ) Pr F ( ) ( m ) q + + ( m + ) ; µ ( m ) q + ( m + ) ; Pr exp σ ( m ) q + + ( m + ) ; µ ( m ) q + m Pr l ( ) ( + ) ; σ m q + + ( m + ) ; ( V W l ) Pr is the upper tolerace factor hvdv ( ) (3) l w (3)

America Joural of Theoretical ad Applied tatistics 6; 5(-): -6 5 ( m + ) q( m + ) ; ( m + ) q + It follows from (8) (7) ad (3) that ( m + ) ; arg hvgwdvdw ( ) ( ) wl /( r) (3) (33) Goodess-of-fit testig It is assumed that x µ ~ f ( x) exp x µ i σ σ (35) the parameters µ ad σ are uow Thus for this example we have that r 5 m 3l 5 95 95 8 ( ) 66 (36) i i It ca be show that the Taig ito accout (3) ad (33) we have (8) This completes the proof Corollary 3 If m the r ( ) l + if l( ) r ( ) l if < l( ) (34) The result similar to that of (34) ca by foud i [7] Remar It will be oted that a upper tolerace limit may be obtaied from a lower tolerace limit by replacig by by 4 Practical Example A idustrial firm has the policy to replace a certai device used at several locatios i its plat at the ed of 4-moth itervals It does t wat too may of these items to fail before beig replaced hipmets of a lot of devices are made to each of three firms Each firm selects a radom sample of l 5 items ad accepts his shipmet if o failures occur before a specified lifetime has accumulated I order to fid this specified lifetime the maufacturer wishes to tae a radom sample of size 5 ad to calculate the lower oe-sided simultaeous tolerace limit () which is expected to capture a certai proportio 95 or more of the populatio of selected items (m3l) with a give cofidece level 95 This tolerace limit is such that oe ca say with a certai cofidece that at least % of the product selected for testig by firms will operate loger tha () The resultig lifetimes (rouded off to the earest moth) of the iitial sample of size 5 from a populatio of the aforemetioed devices are give i Table Table The resultig lifetimes of the iitial sample of size 5 Observatios (i terms of moth itervals) 3 4 5 6 7 8 8 9 4 7 5 9 3 4 5 9 3 35 4 47 54 6 j j+ ( i )( ) + i i j i () j () j+ ( i )( ) + i i i (37) are iid () rv s (Nechval et al []) We assess the statistical sigificace of departures from the two-parameter expoetial model () by performig the Kolmogorov-mirov goodess-of-fit test We use the K statistic (Muller et al []) The rejectio regio for the α5 level of sigificace is {K K ;α } The percetage poits for K ;α were give by Muller et al [] For this example K 8 < K 3;α5 36 (38) Thus there is ot evidece to rule out the two-parameter expoetial model Now the lower oe-sided simultaeous -cotet tolerace limit at the cofidece level () (o the order statistic Y from a set of m future ordered observatios Y Y m ) ca be obtaied from () ice l( ) 5< 9 (39) l ( m + ) q( m + ) ; 989848 ( m + ) q + it follows from () that ( m + ) ; r ( ) 4 (4) (4) tatistical iferece Thus the maufacturer has 95% assurace that o failures will occur i the proportio 95 or more of the populatio of selected items before 4 moth itervals

6 Nicholas A Nechval ad Kostati N Nechval: Tolerace imits o Order tatistics i Future amples Comig from the Two-Parameter Expoetial Distributio 5 Coclusio Tolerace limits ejoy a fairly rich history i the literature ad have a very importat role i egieerig ad maufacturig applicatios I cotrast to other statistical limits commoly used for statistical iferece the tolerace limits (especially for the order statistics) are used relatively rarely Oe reaso is that the theoretical cocept ad computatioal complexity of the tolerace limits is sigificatly more difficult tha that of the stadard cofidece ad predictio limits Thus it becomes ecessary to use the iovative approaches which will allow oe to costruct tolerace limits o future order statistics for may populatios Refereces [] V Medehall A bibliography o life testig ad related topics Biometria vol V pp 5 543 958 [] I Guttma O the power of optimum tolerace regios whe samplig from ormal distributios Aals of Mathematical tatistics vol VIII pp 773 778 957 [3] A Wald ad J Wolfowitz Tolerace limits for a ormal distributio Aals of Mathematical tatistics vol VII pp 8 5 946 [4] W A Wallis Tolerace itervals for liear regressio i ecod Bereley ymposium o Mathematical tatistics ad Probability iversity of Califoria Press Bereley 95 pp 43 5 [5] J K Patel Tolerace limits: a review Commuicatios i tatistics: Theory ad Methodology vol 5 pp 79 76 986 [6] I R Dusmore ome approximatios for tolerace factors for the two parameter expoetial distributio Techometrics vol pp 37 38 978 [7] W C Guether A Patil ad V R R ppuluri Oe-sided -cotet tolerace factors for the two parameter expoetial distributio Techometrics vol 8 pp 333 34 976 [8] M Egelhardt ad J Bai Tolerace limits ad cofidece limits o reliability for the two-parameter expoetial distributio Techometrics vol pp 37 39 978 [9] W C Guether Tolerace itervals for uivariate distributios Naval Research ogistics Quarterly vol 9 pp 39 333 97 [] G J Hah ad W Q Meeer tatistical Itervals: A Guide for Practitioers New Yor: Joh Wiley & os 99 [] N A Nechval ad K N Nechval Characterizatio theorems for selectig the type of uderlyig distributio i Proceedigs of the 7th Vilius Coferece o Probability Theory ad d Europea Meetig of tatisticias Vilius: TEV 998 pp 35 353 [] P H Muller P Neuma ad R torm Tables of Mathematical tatistics eipzig: VEB Fachbuchverlag 979