STATISTICAL INFERENCE IN ACCELERATED LIFE TESTING WITH GEOMETRIC PROCESS MODEL. A Thesis. Presented to the. Faculty of. San Diego State University

Size: px
Start display at page:

Download "STATISTICAL INFERENCE IN ACCELERATED LIFE TESTING WITH GEOMETRIC PROCESS MODEL. A Thesis. Presented to the. Faculty of. San Diego State University"

Transcription

1 STATISTICAL INFERENCE IN ACCELERATED LIFE TESTING WITH GEOMETRIC PROCESS MODEL A Thesis Presented to the Faculty of San Diego State University In Partial Fulfillment of the Requirements for the Degree Master of Science in Statistics by Shan Huang Spring 2011

2

3 iii Copyright c 2011 by Shan Huang

4 iv ABSTRACT OF THE THESIS STATISTICAL INFERENCE IN ACCELERATED LIFE TESTING WITH GEOMETRIC PROCESS MODEL by Shan Huang Master of Science in Statistics San Diego State University, 2011 In reliability analysis, accelerated life testing is the most common way to assess a product s life. Under such test settings, products are tested at higher-than-usual levels of stress to induce early failure. The goal of accelerated life analysis is to utilize the test data to extrapolate a product s life distribution and its associated parameters at a normal stress level. The current study introduces the geometric process model for the analysis of accelerated life testing with an exponential life distribution under constant stress. The geometric process describes a simple monotone process and has been applied to a variety of situations such as the maintenance problems in engineering. By assuming that the lifetime under increasing stress levels forms a geometric process, we derive the maximum likelihood estimators of the exponential parameter under three different censoring schemes: no censoring, Type I and Type II censoring. For each censoring type, we also derive confidence intervals for the parameters using both asymptotic distribution and the parametric bootstrap method. The performance of the estimators is evaluated by a simulation study with different pre-fixed parameters. The study also considers whether the assumption of geometric process is satisfied under a wide-acknowledged log-linear relationship between life and stress. Some simulated numerical examples are presented to illustrate how to pre-design the stress levels so that the geometric process model could be applied. We also use these numerical examples to compare the performance of the proposed geometric model and the traditional failure time regression model. The results of simulation studies, as well as the advantage of the proposed model are summarized in the conclusion part, and the prospect of future works is briefly discussed.

5 v TABLE OF CONTENTS PAGE ABSTRACT... iv LIST OF TABLES... vii ACKNOWLEDGEMENTS... viii CHAPTER 1 INTRODUCTION Introduction Purpose of the Study Format of the Thesis LITERATURE REVIEW Accelerated Life Testing Accelerated Life Model Parameter Estimation The Asymptotic Properties of MLE MLE of Failure-Time Regression Model Geometric Process GP MODEL IN ALT WITH COMPLETE DATA Model Description Estimation Asymptotic Confidence Interval Bootstrap Confidence Interval Simulation Study GP MODEL IN ALT WITH TYPE I CENSORED DATA Model Description Estimation Asymptotic Confidence Interval Bootstrap Confidence Interval Simulation Study GP MODEL IN ALT WITH TYPE II CENSORED DATA... 23

6 vi 5.1 Model Description Estimation Asymptotic Confidence Interval Bootstrap Confidence Interval Simulation Study JUSTIFICATION OF THE GP ASSUMPTION IN ALT Relationship Between Stress and Life Comparison of the Proposed Model and Failure Time Regression Model CONCLUSION BIBLIOGRAPHY APPENDIX R CODE FOR SIMULATION STUDIES... 41

7 vii LIST OF TABLES PAGE Table 1.1. Life Characteristic for Common Life Distributions... 2 Table 3.1. Parameter Estimation for the Complete Simulated Sample with λ = Table 3.2. Parameter Estimation for the Complete Simulated Sample with λ = Table 4.1. Parameter Estimation for Type I Censored Simulated Sample with λ = Table 4.2. Parameter Estimation for Type I Censored Simulated Sample with λ = Table 5.1. Parameter Estimation for Type II Censored Simulated Sample with λ = Table 5.2. Parameter Estimation for Type II Censored Simulated Sample with λ = Table 6.1. Comparison Between the GP and the FTR Model with Complete Data Table 6.2. Comparison Between the GP and the FTR Model with Type I Censored Data Table 6.3. Comparison Between the GP and the FTR Model with Type II Censored Data... 35

8 viii ACKNOWLEDGEMENTS I would like to express my gratitude to my thesis advisor, Professor Jianwei Chen, for his guidance throughout the development of this thesis. His patience, insights and expertise enabled me to develop an understanding of the subject. I offer my sincere thanks to Prof. Juanjuan Fan and Prof. Tao Xie for serving on my advisory committee and for their guidance and cooperation as thesis committee member. Their advice and suggestions were invaluable. I am also grateful for the help and encouragement of my parents and my boyfriend, and I thank all of those who supported me in any respect during the completion of the thesis.

9 1 CHAPTER 1 INTRODUCTION 1.1 INTRODUCTION Life testing is the most direct way to assess the life characteristics of a product, system or component. In traditional life testing and reliability experiments, researchers used to analyze time-to-failure data obtained under normal operating conditions in order to quantify the product s failure-time distribution and its associated parameters. However, such life data is more and more difficult to obtain as a result of the great reliability of today s products, and the small time period between design and release. This problem has motivated researchers to develop new life testing method and obtain timely information on the reliability of product components and materials. Accelerated life testing (ALT) is then adopted and widely used in manufacturing industries. In such testing situations, products are tested at higher-than-usual levels of stress (e.g., temperature, voltage, humidity, vibration or pressure) to induce early failure. The life data collected from such accelerated tests is then analyzed and extrapolated to estimate the life characteristics under normal operating conditions. Three types of stress loadings are usually applied in accelerated life tests: constant stress, step stress and linearly increasing stress. The constant stress loading, which is a time-independent test setting, has several advantages over the time-dependent stress loadings. For example, most products are assumed to operate at a constant stress under normal use. Therefore, a constant stress test mimics actual use. Besides, it is easier to run and to quantify a constant stress test. In the current study, we only discuss the application of constant stress in accelerated life testings. Censoring always occur in accelerated testings. Tests are stopped before all items fail in order to reduce test time and expense. In Type I censoring, items at each stress level start to operate at the same time, and the testing is terminated at a prescribed time t with a random number of observed failures. In Type II censoring, items at each stress level also start to operate at the same time, but the testing is terminated upon the r th failure with a random time t (r). Both of the two censoring schemes are widely applied and discussed, with Type I common in practice and Type II common in theoretical literature. In this study, we discuss the accelerated life testings under three different censoring schemes: no censoring, Type I and Type II censoring.

10 2 Most accelerated test models have two components: a life distribution for a unit s behavior at a constant stress level, and an assumed relationship between the life characteristics and the stress factor. The exponential distribution is one of the most extensively used life distributions in the areas of life testing and reliability. For one reason, it is often a reasonable approximation for the distribution of lifetime for a complex electronic system [22]. For another, it could provide insight into other distributions like the Weibull and lognormal. Other commonly used life distributions include normal, lognormal, Weibull and extreme value distributions. For the sake of simplicity, we assume the underlying life distribution is exponential in our discussion. The relationship between the life distribution (or the life distribution characteristic under consideration) and the stress levels is another essential component of the accelerated test model. It expresses the effect of changing factors like temperature, voltage, humidity on a product s failure time. Most relationships are formulated with a general log-linear expression between the life characteristic and the (possibly transformed) accelerating factors [23]. The selection of the life characteristic always depends on the assumed underlying life distribution. For example, when considering the exponential distribution, the mean life is chosen to be the life characteristic that is stress dependent. When considering the lognormal distribution, the median life is the typical life characteristic. Each specified accelerated life model relates life characteristic to the explanatory variable (or stress factor) in some particular way. A more detailed discussion of life and stress relationship will be given in Chapter 6. Table 1.1 lists some common life distributions and the corresponding life characteristics. Table 1.1. Life Characteristic for Common Life Distributions Distribution Notation Density Function Life Characteristic Meaning Exponential Exp(λ) f(x λ) = λe λx, x 0 1 λ Mean life Weibull W (α, β) f(x α, β) = β α β x β 1 e ( x α )β, x 0 α 63.2% life Lognormal LN(µ, σ 2 ) f(x µ, σ) = 1 2πσx e 1 2σ 2 (log x µ)2, x 0 e µ Median life Once a life distribution and a life-stress relationship have been selected, the distribution parameters that govern the life characteristics need to be determined. Various statistical methods have been applied to estimate the parameters and their confidence limits of an accelerated testing model; two of them are generally used by researchers: graphical estimation and maximum likelihood estimation. Graphical method is a simple and easy way to perform the estimation. It first estimates the parameters of the life distribution at each individual stress level, and then plots the life

11 3 characteristic vs. stress on a paper that linearizes the assumed life-stress relationship. The parameters of the life-stress relationship are then estimated from the plot. Clearly, the graphical method treats the life distribution and the life-stress relationship separately. One of the drawbacks of this method is that in some problems, lots of the life data are censored; there is even no failure occurs at lower stress levels. In such situations, the standard techniques of graphical estimation for the parameters cannot be used, since the values of the censored observations are not known. Maximum likelihood method is a better approach for fitting models to censored data. Using the MLE method, the life distribution and the life-stress relationship can be treated as one complete model that describes both. The life data that are observed to have an exact failure time and are censored to have a partially known value are included in the likelihood function. As we discuss in Chapter 2, a maximum likelihood estimator has some good properties including the asymptotic normality. Since the MLE method is straightforward, and applies to most theoretical models and censored data, researchers prefer to use it when fitting an accelerated testing model. 1.2 PURPOSE OF THE STUDY In this study, we introduce the geometric process model for the analysis of accelerated life testing with different types of censoring and an exponential life distribution under constant stress. The geometric process is a simple monotone process introduced by Lam [11, 12] and has been applied to a wide field of maintenance problems in engineering. In brief, a geometric process describes a stochastic process {X n, n = 1, 2,... where there exists a real-valued a > 0 such that{a n 1 X n, n = 1, 2,... forms a renewal process. We have reason to believe that in an accelerated testing, lifetime will be stochastically decreasing with respect to increasing stress levels. Therefore, the geometric process model is a natural approach to study such a problem. Suppose {X k, k = 1, 2,..., s are lifetimes under each stress level and X 1 follows an underlying life distribution Exp(λ). We assume that {X k, k = 1, 2,..., s is a geometric process with ratio a. It can be shown that the life distribution at k th stress level is Exp(a k 1 λ). In that case, the likelihood function of the exponential geometric process contains only two parameters: a and λ. The objective of this study is to develop inference for those two parameters of the corresponding geometric process model by the maximum likelihood estimation. The current study also concerns with the statistical properties of maximum likelihood estimators of the model parameters. The asymptotic equivalence of ML estimators based on censored data have been discussed by Escobar and Meeker [8]and Bhattachary [4] (please refer to Chapter 2 for a detailed review). We derive the asymptotic confidence intervals for

12 4 estimators of the geometric process model under the accelerated testing framework, and compare them with the CIs that are obtained by the parametric bootstrap method. Simulation experiments are also conducted based on three types of censored data to assess how the asymptotic theory applies with small to moderate sample sizes as in practice. 1.3 FORMAT OF THE THESIS The rest of the thesis is organized as follows. Chapter 2 reviews the literature related to models and methodologies used in accelerated life tests, as well as theories and applications of geometric process developed since Chapter 3 to Chapter 5 discuss the analyses of accelerated life testing with geometric process model based on complete, Type I censored and Type II censored data, respectively. In each of the three chapters, we describe the proposed model, present the maximum likelihood estimates of the parameters and then derive their asymptotic and bootstrap confidence intervals. We also evaluate the performance of confidence intervals in terms of probability coverages through a simulation study. Chapter 6 briefly describes the log-linear relationship between a life characteristic and the stress level. It also discusses how the assumption of geometric process is satisfied under this relationship. Two common life-stress relationship models: Arrhenius and Inverse power models are presented to illustrate how to pre-fix stress levels to fulfill the geometric process assumption. A numerical example compares the geometric process model and the traditional failure time regression model under certain stress levels. Chapter 7 contains our concluding remarks of the current study and a prospect of further researches.

13 5 CHAPTER 2 LITERATURE REVIEW 2.1 ACCELERATED LIFE TESTING Accelerated test experiments have been used in manufacturing industries for many decades. A vast and growing literature on statistical methods for accelerated life testing could be found since the late 1960 s. The main purpose of the statistical researches is to utilize test data that obtained under high stress levels to estimate the life of the product at lower stress levels encountered in normal use. In this section, we introduce the most commonly used accelerated life model and the maximum likelihood method for parametric estimation. The asymptotic properties of the ML estimators are discussed, and some research trends in the statistical inference of ALT are summarized Accelerated Life Model A simple, commonly used model that characterizes the effect of explanatory variables on lifetime is failure-time regression model. With this model, the life time is assumed to follow some distribution such as the exponential, the Weibull or the lognormal distribution. Given the assumed distribution, the accelerated life model is described by the dependence of distribution parameter on the applied stresses. For example, if the life time is described by a Weibull distribution with scale parameter α and shape parameter β, typically for parametric regression model, the parameter α is then modeled as a function of covariates and the parameter β is held fixed (see [10]). As with the model of exponential distribution f(x λ) = λe λx, which is a special case of Weibull distribution, 1/λ will be reparameterized with stress variables and regression coefficients. In general, if the life time is described by a stochastic model f(x θ) with parameter θ, then the life characteristic, denoted by g(θ), can be expressed as a function of the independent variables and the corresponding coefficients: g(θ) = φ(z; γ 0, γ 1,...), (2.1) where z is derived covariates; φ(.) is a specified functional form. Most published case studies of accelerated life testing choose the form φ(.) = exp(.), since it takes only positive values and thus guarantees θ > 0 for all possible z (see [10]). In that case, equation (2.1) could be

14 6 written as log [g(θ)] = γ 0 + γ 1 z, (2.2) if a single stress factor is applied in the life testing. This relationship indicates that the logarithm of a life characteristic is a linear function of the stress-related covariates. In particular, as Nelson [22] stated in his book, for the exponential life distribution with single stress, the mean life 1/λ is log 1 λ = γ 0 + γ 1 z. (2.3) Parameter Estimation Statisticians generally prefer the method of maximum likelihood to obtain parameter estimates when the life data are censored. In the current section, we first review the likelihood construction and the asymptotic theories for ML estimates applied to censored data, and then extend the discussion to failure-time regression models where explanatory variable (in our study, the stress factor) is incorporated. The construction of likelihood function of censored data has been discussed in numerous textbooks and journal articles (see [1, 7, 10, 22] for example). In a typical life test, n items are placed under observation and each failure time is recorded. The test terminates either at a pre-determined fixed time t (Type I censor) or after a fixed number of sample items fail (Type II censor). Let f(x) be the PDF and F (x) the CDF for chosen life distribution with parameter θ, and (n r) items survived beyond the time of termination. In both type I and type II censoring, the likelihood function may be written as [ r ] n! L(θ) = f(x (i) ) [1 F (x T )] n r, (2.4) (n r)! i=1 where in type I censoring, the time of termination x T = t, and in type II censoring x T = x (r). Associated with the likelihood function is the efficient score vector U(θ) = ln L(θ), θ and the maximum likelihood estimator ˆθ of the parameter θ verifies the equation: to maximize the likelihood function (2.4). U(ˆθ) = 0

15 2.1.3 The Asymptotic Properties of MLE 7 An important advantage of the maximum likelihood method is that under certain regularity conditions (usually met in practice), ML estimators are asymptotically normal [22]. The asymptotic variance-covariance matrix is obtained by inverting the Fisher information matrix, which is defined by [ ] 2 I(θ) = E θ ln L(θ). θ2 The asymptotic theory of ML can not be directly applied in the Type II censoring scheme, since lifetime are not independent due to the pre-fixed number of failures. Bhattachary [4] provided some procedures to establish the limiting normality and uniform strong convergence of ML estimators, and showed that using certain sets of regularity conditions, maximum likelihood estimation of parameters from Type II censored samples are consistent, asymptotically normally distributed and efficient. Escobar and Meeker [8] studied the case of location-scale distributions, and further showed that although the Fisher information matrices for Type I and Type II censoring proceed differently, under standard regularity conditions, the information matrices obtained are asymptotically equivalent MLE of Failure-Time Regression Model When stress factors are incorporated, as we discussed earlier in this Chapter, θ can be expressed as an assumed function of the stress factors z and their coefficients γ 0, γ 1,.... Under a certain stress level z k, the distribution parameter θ k is θ k = θ k (z k ; γ 0, γ 1,...). Consequently, the likelihood function in (2.4) can be written as [ rk ] n k! L(θ k ) = L(γ 0, γ 1,... x k ) = f(x k(i) ) [1 F (x kt )] n k r k. (2.5) (n k r k )! The analysis reduces to the inferences of γ 0, γ 1,... from life tests and the estimation of the parameter θ at normal stress level. The maximum likelihood estimates ˆγ 0, ˆγ 1,... are the coefficient values that maximize the sample log likelihood lnl(γ 0, γ 1,...), and are usually obtained by iterative numerical optimization methods (see Nelson [22]). According to the asymptotic theory of MLE, the joint sampling distribution of ˆγ 0, ˆγ 1,... converges to a joint normal distribution. Nelson [22] presented in detail the calculation of sample likelihood of a i=1 set of data and the ML estimates of model coefficients. In accelerated life studies, researchers are more concerned with the life distribution under the normal operating condition. The life characteristics g(θ) are functions of the

16 8 regression parameters γ 0, γ 1,.... Therefore, the ML estimates ˆθ of θ are ˆθ = g 1 φ( ˆγ 0, ˆγ 1,...). The approximate variance and confidence intervals for θ are obtained by delta method based on the asymptotic property of functions of ML estimators. A detailed illustration could be found in [10, 20, 22]. The failure-time regression model with maximum likelihood method has been extensively studied in the literature. Evans [9] considered the exponential distribution with λ related to stress by the Arrhenius equation, which is a special case of the log-linear relationship. He derived the ML estimates of λ at specified stress and used asymptotic theory to obtain the confidence limits. McCool [17] treated a 2-parameter Weibull distribution with the scale parameter a power function of stress. A numerical scheme was applied to solve the nonlinear simultaneous equations for the ML estimates of the shape parameter and the stress-life exponent. Watkins [25] discussed how to obtain MLE of parameters in a Weibull log-linear model and illustrated the numerical approach by two examples. More recently, Watkins [24] studied a variant of the basic Type II censoring scheme, and considered the statistical properties of maximum likelihood estimators of Weibull log-linear model under the certain design. Despite the popularity and wide application of the failure-time regression model, researchers are still seeking development spaces for the statistical methods of accelerated life test. Watkins [26] argued that even though the log-linear function is just a simple reparameterization of the original parameter θ of the life distribution, from a statistical perspective, it is preferable to work with the original parameters. Inspired by Watkins, Balakrishnan et al [2], and Balakrishnan and Xie [3] considered the exponential lifetimes under the simple step-stress experiment terminated by Type II censoring and Type I hybrid censoring schemes, respectively. Both articles implemented the maximum likelihood method to directly estimate the scale parameter θ 1, θ 2 at each of the two stress levels. Different types of confidence intervals are also derived and compared by simulation studies. Their studies are limited to two stress levels since a model with more stress levels demands an intensive computation. Although the accelerated life test with step-stress is beyond the scope of the present discussion, the above studies provide a new perspective of modeling the accelerated life data with an underlying exponential distribution. Instead of developing inferences for the parameters γ 0 and γ 1 of the log-linear link function, it is recommended to derive the MLEs of the unknown parameters of life distribution directly, and construct their confidence intervals. In the present study, we introduce the geometric process model to study the lifetime data

17 under s constant stress levels with different censoring schemes, and develop inferences of the original life parameter in a simple yet precise way GEOMETRIC PROCESS The concept of geometric process is first introduced by Lam in 1988 [11, 12] when he studied the problem of repair replacement. The geometric process is defined as follows (see Lam [14]): A stochastic process {X n, n = 1, 2,... is said to be a Geometric Process (GP), if there exists a real-valued a > 0 such that {a n 1 X n, n = 1, 2,... forms a renewal process. The positive number a is called the ratio of the GP. It is clear to see that a GP is stochastically increasing if 0 < a < 1; it is stochastically decreasing if a > 1. Therefore, the GP is a natural approach to analyze data from a series of events with trend. It can be shown that if {X n, n = 1, 2,... is a GP and the p.d.f. of X 1 is f with mean λ and variance σ 2, then the p.d.f. of X n will be given by a n 1 f(a n 1 x), with E(X n ) = λ/a n 1 and V ar(x n ) = σ 2 /a 2(n 1). Thus, a, λ and σ 2 are three important parameters of a GP. In our case of the exponential distribution, two parameters will be estimated, a and λ. The statistical inference for geometric process is studied in both nonparametric and parametric ways. Lam introduced least square and modified moment estimation of parameters for GP, and studied the asymptotic normal properties of these estimators (see [14]). Lam and Chan [15] derived the maximum likelihood estimators of a, λ and σ 2 of the GP with lognormal distribution. They showed that the maximum likelihood estimators are more efficient than the nonparametric estimators. Chan, Lam, and Leung [5] estimated the parameters of the GP with Gamma distribution by parametric methods including maximum likelihood method along with some nonparametric methods previously proposed by Lam. They also illustrated how to derive the limiting distribution of the ML estimators and construct the confidence intervals. Chen [6] provided the Bayesian approach to the estimation of parameters in a GP with several popular life distributions including the exponential and lognormal distributions. The comparison between the Bayesian estimators and ML estimators is also examined under simulation studies. The geometric process model is first applied to investigate a repair replacement model for a one-unit deteriorating system, in which the successive survival times of the system are stochastically non-increasing, while the consecutive repair times of the system after failure are stochastically non-decreasing (see [11, 12]). Since then, a lot of research work on different GP maintenance models has been done. For example, Lam and Zhang [16] introduced the geometric process model in the analysis of a two-component series system with one

18 10 repairman. Lam [13] studied the geometric process model for a multistate system and determined an optimal replacement policy to minimize the long-run average cost per unit time. Zhang [28] used the geometrical process to model a simple repairable system with delayed repair. Large amount of studies in maintenance problems and system reliability have shown that a geometric process model is a good and simple model for analysis of data with a single trend or multiple trends. So far, there is no study that utilizes the geometric process in the analysis of accelerated life test with censored data. The current study aims to introduce the geometric process model to analyze a series of life data that obtained from several increasing stress levels. Statistical inference of the parameters in the GP will be made and examined through a simulation study.

19 11 CHAPTER 3 GP MODEL IN ALT WITH COMPLETE DATA 3.1 MODEL DESCRIPTION Suppose that we are given an accelerated life test with s increasing stress levels. A random sample of n identical items are placed under each stress level and start to operate at the same time. Whenever an item fails, it will be removed from the test. We denote by k the index for stress level; k = 1, 2,..., s, and denote by i the index for test item under each stress; i = 1, 2,..., n. Each item s observed failure time can be written as x ki. The geometric model for accelerated life test is based on two assumptions: Assumption 1. Lifetime at the design stress level follows a certain distribution. In the current study, it is assumed to be exponential with failure rate λ. The probability density function (PDF) is given by { λe λx if x > 0, f X (x λ) = 0 if x 0. Assumption 2. Let the sequence of random variables X 0, X 1,..., X s denote the lifetimes under each stress level, where X 0 denotes item s lifetime under the design stress at which items will operate ordinarily. We assume {X k, k = 0, 1, 2,... s is a geometric process with ratio a > 0. Based on the definition of geometric process [14], if the density function of X 0 is f, then the probability density function of X k will be given by: a k f(a k x), k = 0, 1, 2,..., s. (3.1) In our case of exponential distribution, the PDF of a test item at the k th stress level is: { a k λe ak λx if x > 0, f Xk (x λ) = 0 if x 0. (3.2) The model in (3.2) indicates that if lifetime under a sequence of increasing stress levels form a geometric process with ratio a, and if the life distribution at the design stress level is exponential with failure rate λ, then the life distribution at the k th stress level is also exponential with failure rate a k λ. This conclusion and the above mentioned assumptions will also be utilized in other censored schemes described in Chapter 4 and Chapter 5.

20 ESTIMATION In accelerated life test, lifetime under the design stress level will not be observed. Therefore, based on the observed data in a total s stress levels, the likelihood function for the exponential geometric process is given by s n L(a, λ x 1,..., x s ) = a k λ exp{ λa k x ki = = k=1 i=1 s n {a kn λ exp{ λa k x ki (3.3) k=1 i=1 s n {a kn λ n exp{ λa k x ki. k=1 i=1 It follows that l(a, λ) = = { s n log a kn λ n exp{ λa k x ki k=1 i=1 { s kn log a + n log λ λa k k=1 i=1 n x ki. (3.4) The first and second derivatives of l(a, λ) are: s l(a, λ) = a s l(a, λ) = λ 2 s l(a, λ) = a2 2 l(a, λ) = 2 a λ k=1 k=1 k=1 { kn n a λ( x ki )ka k 1, { n λ ak ( { i=1 n x ki ), i=1 kn λk(k 1)ak 2 a2 n s λ a l(a, λ) = k=1 i=1 n x ki, i=1 ka k 1 x ki, 2 l(a, λ) = sn λ2 λ. (3.5) 2 The MLEs of a and λ exist but do not have closed forms. The Newton Iterative Method is applied to obtain â and ˆλ. Below is the algorithm of Newton Iterative Method: ( ) (0) a 1. Start with an initial value of. λ

21 13 2. For the (t + 1) th iteration, t 0, ( a λ ) (t+1) = ( a λ 3. Repeat (2) until converge. ) (t) 2 l(a, λ) ( 2 l(a, λ) a 2 a λ 2 l(a, λ) 2 l(a, λ) λ a λ 2 ) 1 ( a (a,λ) (t) λ l(a, λ) l(a, λ) ) (a,λ) (t). 3.3 ASYMPTOTIC CONFIDENCE INTERVAL Let I(a, λ) denote the Fisher information matrix ] of a and λ. The observed Fisher information matrix of a and λ is Î(a, λ) = [ Î11 Î 12 Î 21 Î 22 [ ] 2 Î 11 = l(a, λ) = a2 = ns(1 + s) 2a 2 + λ [ 2 Î 12 = Î21 = [ 2 Î 22 = l(a, λ) λ2 s k=1 l(a, λ) a λ ] a=â,λ=ˆλ, where { s kn + λk(k 1)ak 2 a2 k=1 { n k(k 1)a k 2 x ki = sn λ 2. ] = s k=1 i=1 n ka k 1 x ki, i=1 n x ki, i=1, (3.6) The variances of â and ˆλ can be obtained through the observed information matrix as [ ] [ ] 1 V ar(â) I11 ˆ =. V ar(ˆλ) Iˆ 22 The 100(1 α)% asymptotic confidence intervals for a and λ are then given by [ ] â ± Z 1 α 2 V ar(â), and [ ˆλ ± Z 1 α 2 V ar(ˆλ) ], respectively. 3.4 BOOTSTRAP CONFIDENCE INTERVAL We construct the confidence intervals for a and λ based on the parametric bootstrap, using the percentile bootstrap interval method. We describe the algorithm to obtain the coverage for bootstrap CI as below. 1. Based on the original sample X ki, obtain â and ˆλ, the MLEs of a and λ, by Equation (3.5) and the Newton iterative method.

22 14 2. For b = 1, 2,..., B, based on â and ˆλ, simulate a bootstrap sample Xki, k = 1, 2,..., s; i = 1, 2,..., n, where X ki Exp(x; âkˆλ). 3. Compute â (b) and ˆλ (b), the MLEs of a and λ, by Equation (3.5) and the Newton iterative method. 4. Repeat steps 2-3 for B times and obtain â (1), ˆλ (1), â (2), ˆλ (2),..., â (B), ˆλ (B). 5. The bootstrap percentile confidence interval endpoints for a and λ are the α/2 and 1 α/2 quantiles of â (1),..., â (B) and ˆλ (1),..., ˆλ (B), respectively. 3.5 SIMULATION STUDY To evaluate and compare the performance of the methods of inference described in the preceding section, we conduct a simulation study. The R code is listed in the Appendix. We first generate a sample X ki, k = 1, 2,..., s; i = 1, 2,..., n, where X ki Exp(x; a k λ). We choose the values of the parameters a = 1.03, 1.05, 1.07 and λ = 0.5, The ratios a are chosen to be close to 1 since the decreasing trend of lifetime in practice is usually not pronounced. The number of stress levels is chosen to be s = 4 or 6; the number of test products at each stress level is n 1 = = n s = 10 or 20. The estimators and the corresponding summary statistics are obtained by our proposed model and the Newton iteration method. For a given sample with different choices of a and λ, we list the average of the ML estimations (mean), the sample standard deviation of the estimates (SE), the average of asymptotic standard error ( ˆ SE), the square root of mean squared error ( MSE) and the coverage rate of the 95% confidence interval for a and λ. The CIs are obtained by two methods: the asymptotic distribution and the parametric bootstrap. The numerical results presented in Table 3.1 and 3.2 are based on 500 simulations and 500 bootstrap replications. Table 3.1 and 3.2 summarizes the results of the estimates for a and λ. We observe that â and ˆλ estimates the true parameters a and λ quite well with relatively small mean squared errors. The estimated standard error also approximates well the sample standard deviation of the 500 estimates. For a fixed a and λ, we compare their estimates across the four different cases and find that as n and s decreases, the mean squared errors of â and ˆλ get larger. This may be because that a larger sample size results in a better large-sample approximation for the distribution of â and ˆλ, so the inference for the parameters is more precise. We also notice that the coverage probabilities of the asymptotic confidence interval are close to the nominal level and do not change much across the four cases. The parametric bootstrap confidence intervals have similar coverage rates as the approximate CIs. This result indicates that the proposed model and the asymptotic approximation work well under the situation where no censoring occurs.

23 15 Table 3.1. Parameter Estimation for the Complete Simulated Sample with λ = 0.5 a Estimator M ean SE MSE case 1: n = 20, s = 6 ˆ SE 95% CI Coverage Approx Bootstrap 1.03 â ˆλ â ˆλ â ˆλ case 2: n = 10, s = â ˆλ â ˆλ â ˆλ case 3: n = 20, s = â ˆλ â ˆλ â ˆλ case 4: n = 10, s = â ˆλ â ˆλ â ˆλ

24 16 Table 3.2. Parameter Estimation for the Complete Simulated Sample with λ = 0.25 a Estimator M ean SE MSE case 1: n = 20, s = 6 ˆ SE 95% CI Coverage Approx Bootstrap 1.03 â ˆλ â ˆλ â ˆλ case 2: n = 10, s = â ˆλ â ˆλ â ˆλ case 3: n = 20, s = â ˆλ â ˆλ â ˆλ case 4: n = 10, s = â ˆλ â ˆλ â ˆλ

25 17 CHAPTER 4 GP MODEL IN ALT WITH TYPE I CENSORED DATA 4.1 MODEL DESCRIPTION Suppose that there are s increasing stress levels and under each level n items are inspected. We denote by k the index for stress level; k = 1, 2,..., s, and denote by i the index for test item under each stress; i = 1, 2,..., n. In the Type I censored scheme, the test at each stress level terminates at time t. An item s exact failure time is observed only if its lifetime x ki t. Assume at the k th stress level we observe r k failures before the test is suspended, 0 r k n. Correspondingly, (n r k ) units survive the entire test without failing. The observed ordered failure times under the k th stress level can be written as x k(1) x k(2)... x k(rk ). Note that t is fixed in advance and r k is random. The two assumptions of the geometric process model in accelerated life testing still need to be held. As is described in Chapter 3, the PDF of a test item at the k th stress level is: { a k λe ak λx if x > 0, f Xk (x) = (4.1) 0 if x 0. Consequently, the CDF of a test item at the k th stress level is: F Xk (x) = x 0 a k λe ak λx dx (4.2) = 1 e ak λx. The probability that an item censored at time t is: S Xk (t) = 1 F Xk (t) = e ak λt. (4.3) For those (r k ) items that fail at time t, their order statistics can be denoted by X k(i) with PDF: f Xk(i) (x) = n! (i 1)!(n i)! f X k (x)[f Xk (x)] i 1 [1 F Xk (x)] n i, (4.4) i = 1, 2,..., r k. Given (4.3) and (4.4), the likelihood function of observed data at the k th stress level can be expressed as L k = n! (n r k )! f X(x k(1) )... f Xk (x k(rk ))[S Xk (t)] n r k. (4.5)

26 ESTIMATION In this section, we derive the ML estimates of a and λ from the likelihood function presented in (4.5). The substitution of (4.1) and (4.3) into (4.5) yields { ( rk ) n! L k (a, λ) = (n r k )! (λak ) r k exp λa k x k(i) + (n r k )t, i=1 0 x k(1)... x k(rk ) t. (4.6) It follows that the likelihood function of observed data in a total s stress levels is: L(a, λ) = L 1 L 2 L s { { ( s rk ) n! = (n r k )! (λak ) r k exp λ x k(i) + (n r k )t, k=1 Then it can be shown that l(a, λ) = 0 x k(1)... x k(rk ) t; 1 k s. (4.7) { s ( ) ( rk ) n! log + r k log(λa k ) λa k x k(i) + (n r k )t. (4.8) (n r k )! k=1 The first and second derivatives of l(a, λ) are: s l(a, λ) = a s l(a, λ) = λ 2 s l(a, λ) = a2 2 l(a, λ) = 2 a λ 2 s l(a, λ) = λ2 k=1 k=1 k=1 { kr k a { r k λ ak { kr k a 2 λkak 1 s l(a, λ) = λ a k=1 i=1 i=1 ( rk ) x k(i) + (n r k )t, i=1 ( rk ) x k(i) + (n r k )t, i=1 λk(k 1)ak 2 k=1 ( rk ) x k(i) + (n r k )t, i=1 { ( rk ) ka k 1 x k(i) + (n r k )t, i=1 { r k λ 2. (4.9) The MLEs of a and λ exist but do not have closed forms. The Newton Iterative Method is applied to obtain â and ˆλ. Below is the algorithm of Newton Iterative Method:

27 19 1. Start with an initial value of ( a λ 2. For the (t + 1) th iteration, t 0, ( ) (t+1) ( a a = λ λ 3. Repeat (2) until converge. ) (0). ) (t) ( 2 a 2 l(a, λ) 2 l(a, λ) l(a, λ) λ 2 a λ 2 l(a, λ) 2 λ a ) 1 ( a (a,λ) (t) λ l(a, λ) l(a, λ) ) (a,λ) (t). 4.3 ASYMPTOTIC CONFIDENCE INTERVAL Let I(a, λ) denote the Fisher information matrix] of a and λ. The observed Fisher information matrix of a and λ is Î(a, λ) = [ Î11 Î 12 Î 21 Î 22 [ ] 2 Î 11 = l(a, λ) = a2 [ 2 Î 12 = Î21 = [ 2 Î 22 = l(a, λ) λ2 s k=1 l(a, λ) a λ ] s = k=1 { kr k ] = a 2 a=â,λ=ˆλ + λk(k 1)ak 2, where ( rk ) x k(i) + (n r k )t, i=1 { ( s rk ) ka k 1 x k(i) + (n r k )t, k=1 i=1 { rk λ 2. (4.10) The variances of â and ˆλ can be obtained through the observed information matrix as [ ] [ ] 1 V ar(â) I11 ˆ =. V ar(ˆλ) Iˆ 22 The 100(1 α)% asymptotic confidence intervals for a and λ are then given by [ ] â ± Z 1 α 2 V ar(â), and [ ˆλ ± Z 1 α 2 V ar(ˆλ) ], respectively. 4.4 BOOTSTRAP CONFIDENCE INTERVAL We construct the confidence intervals for a and λ based on the parametric bootstrap, using the percentile bootstrap interval method. We describe the algorithm to obtain the coverage for bootstrap CI as below: 1. Arrange the original sample X ki in ascending order and find r k such that x k(rk ) t and x k(rk +1) > t. Obtain â and ˆλ, the MLEs of a and λ, by Equation(4.9) and the Newton iterative method.

28 20 2. For b = 1,..., B, based on â and ˆλ, simulate a bootstrap sample Xki, k = 1, 2,..., s; i = 1, 2,..., n, where X ki Exp(x; âkˆλ). 3. Find r k such that x k(r k ) t and x k(r k +1) > t. 4. Compute â (b) and ˆλ (b), the MLEs of a and λ, by Equation (4.9) and the Newton iterative method. 5. Repeat steps 2-4 for B times and obtain â (1), ˆλ (1), â (2), ˆλ (2),..., â (B), ˆλ (B). 6. The bootstrap percentile confidence interval endpoints for a and λ are the α/2 and 1 α/2 quantiles of â (1),..., â (B) and ˆλ (1),..., ˆλ (B), respectively. 4.5 SIMULATION STUDY To evaluate and compare the performance of the methods of inference described in the preceding section, we conduct a simulation study. We first generate a sample X ki, k = 1, 2,..., s; i = 1, 2,..., n, where X k Exp(x; a k λ). Similar to the simulation study in Chapter 3, we choose the values of the parameters a = 1.03, 1.05, 1.07 and λ = 0.5, The number of stress levels is chosen to be s = 6; the number of test products at each stress level is n = 10 or 20. The termination time t is set to be half or three quarters of the expected lifetime. Therefore, for λ = 0.5, we choose t = 1 or 1.5; for λ = 0.25, we choose t = 2 or 3. In each case, we compare every x ki in the generated sample with termination time t, record the minimum one as the actual observed failure time, and sort them in ascending order as x k(1)... x k(rk ). The estimations and the corresponding statistics are obtained by our proposed model and the Newton iteration method. For different choices of a and λ in the four cases, we list the average of the ML estimations (mean), the sample standard deviation of the estimates (SE), the asymptotic standard error ( ˆ SE), the square root of mean squared error ( MSE) and the coverage rate of the 95% confidence interval for a and λ. The CIs are obtained by two methods: the asymptotic distribution and the parametric bootstrap. The numerical results presented in Table 4.1 and 4.2 are based on 500 simulations and 500 bootstrap replications. From the tables, we observe that the proposed estimators â and ˆλ perform well in all the cases studied. The estimation of a is more accurate than that of λ, as seen by its negligible bias. The proposed estimator ˆλ is consistently larger than the true value of λ, which means we tend to underestimate the mean lifetime of the product at the normal stress level. The estimated standard error approximates well the sample standard deviation of the 500 estimates, with the 95% confidence interval coverage rate close to the nominal level. When the sample size is fixed, the mean squared errors of the estimators decrease as the termination time t gets larger. This is quite expected because large values of t would lead to more failures before the test ends, and thus increase the efficiency of the estimators.

29 21 Table 4.1. Parameter Estimation for Type I Censored Simulated Sample with λ = 0.5 a Estimator M ean SE MSE case 1: n = 20, t = 1, s = 6 ˆ SE 95% CI Coverage Approx Bootstrap 1.03 â ˆλ â ˆλ â ˆλ case 2: n = 20, t = 1.5, s = â ˆλ â ˆλ â ˆλ case 3: n = 10, t = 1, s = â ˆλ â ˆλ â ˆλ case 4: n = 10, t = 1.5, s = â ˆλ â ˆλ â ˆλ

30 22 Table 4.2. Parameter Estimation for Type I Censored Simulated Sample with λ = 0.25 a Estimator M ean SE MSE case 1: n = 20, t = 2, s = 6 ˆ SE 95% CI Coverage Approx Bootstrap 1.03 â ˆλ â ˆλ â ˆλ case 2: n = 20, t = 3, s = â ˆλ â ˆλ â ˆλ case 3: n = 10, t = 2, s = â ˆλ â ˆλ â ˆλ case 4: n = 10, t = 3, s = â ˆλ â ˆλ â ˆλ

31 23 CHAPTER 5 GP MODEL IN ALT WITH TYPE II CENSORED DATA 5.1 MODEL DESCRIPTION Suppose that there are s stress levels and under each level n items are inspected. Let the random variable X ki denote the lifetime of the tested product, where k is the index for stress level and i is the index for test item under each stress. In the Type II censoring scheme, the test at each stress level terminates after r failures are observed. Under the k th stress level, only the r smallest observations of the sample will be collected, whose failure time can be written as x k(1) x k(2)... x k(r). Note that r is specified in advance; the test ends at the r th failure time x k(r), and (n r) units have survived at each stress level. The assumptions of the geometric process model described in Chapter 3 imply that the PDF of a test item at the k th stress level is: { a k λe ak λx if x > 0, f X (x) = 0 if x 0. Consequently, the CDF of a test item at the k th stress level is: F X (x) = x 0 (5.1) a k λe ak λx dx (5.2) = 1 e ak λx. In Type II censoring, the information available consists of the first r order statistics at each stress level. The PDF of the order statistic X k(i) is f X(i) (x) = n! (i 1)!(n i)! f X(x)[F X (x)] i 1 [1 F X (x)] n i, (5.3) i = 1, 2,..., r. For those (n r) units that still alive when the test terminates at each stress level, all we know under Type II censoring is that each one s lifetime exceeds x k(r). The probability that an item censored at time x k(r) is 1 p(x x k(r) ) = e ak λx k(r). (5.4) Both (5.3) and (5.4) would be used in the construction of likelihood function.

32 ESTIMATION stress level is: According to equation (2.4), the likelihood function of the observed test data at the k th L k (a, λ) = { ( r ) n! (n r)! (λak ) r exp λa k x k(i) + (n r)x k(r), i=1 0 x k(1)... x k(r). (5.5) It follows that the likelihood function of observed data in a total s stress levels is { { ( s r ) n! L(a, λ) = (n r)! (λak ) r exp λa k x k(i) + (n r)x k(r), i=1 k=1 0 x k(1)... x k(r) ; 1 k s. (5.6) Then it can be shown that { s ( ) ( r ) n! l(a, λ) = log + r log(λa k ) λa k x k(i) + (n r)x k(r). (5.7) (n r)! i=1 k=1 The first and second derivatives of l(a, λ) are: { ( s r ) kr l(a, λ) = a a λkak 1 x k(i) + (n r)x k(r), k=1 i=1 { ( s r ) r l(a, λ) = λ λ ak x k(i) + (n r)x k(r), k=1 i=1 { ( 2 s r ) l(a, λ) = kr λk(k 1)ak 2 x a2 a2 k(i) + (n r)x k(r), 2 s l(a, λ) = λ2 2 s l(a, λ) = a λ k=1 k=1 k=1 { r λ 2, { ka k 1 ( r i=1 i=1 x k(i) + (n r)x k(r) ) Again, the Newton Iterative Method is applied to obtain â and ˆλ.. (5.8)

Constant Stress Partially Accelerated Life Test Design for Inverted Weibull Distribution with Type-I Censoring

Constant Stress Partially Accelerated Life Test Design for Inverted Weibull Distribution with Type-I Censoring Algorithms Research 013, (): 43-49 DOI: 10.593/j.algorithms.01300.0 Constant Stress Partially Accelerated Life Test Design for Mustafa Kamal *, Shazia Zarrin, Arif-Ul-Islam Department of Statistics & Operations

More information

Statistical Inference on Constant Stress Accelerated Life Tests Under Generalized Gamma Lifetime Distributions

Statistical Inference on Constant Stress Accelerated Life Tests Under Generalized Gamma Lifetime Distributions Int. Statistical Inst.: Proc. 58th World Statistical Congress, 2011, Dublin (Session CPS040) p.4828 Statistical Inference on Constant Stress Accelerated Life Tests Under Generalized Gamma Lifetime Distributions

More information

Step-Stress Models and Associated Inference

Step-Stress Models and Associated Inference Department of Mathematics & Statistics Indian Institute of Technology Kanpur August 19, 2014 Outline Accelerated Life Test 1 Accelerated Life Test 2 3 4 5 6 7 Outline Accelerated Life Test 1 Accelerated

More information

A New Two Sample Type-II Progressive Censoring Scheme

A New Two Sample Type-II Progressive Censoring Scheme A New Two Sample Type-II Progressive Censoring Scheme arxiv:609.05805v [stat.me] 9 Sep 206 Shuvashree Mondal, Debasis Kundu Abstract Progressive censoring scheme has received considerable attention in

More information

Exact Inference for the Two-Parameter Exponential Distribution Under Type-II Hybrid Censoring

Exact Inference for the Two-Parameter Exponential Distribution Under Type-II Hybrid Censoring Exact Inference for the Two-Parameter Exponential Distribution Under Type-II Hybrid Censoring A. Ganguly, S. Mitra, D. Samanta, D. Kundu,2 Abstract Epstein [9] introduced the Type-I hybrid censoring scheme

More information

Statistics - Lecture One. Outline. Charlotte Wickham 1. Basic ideas about estimation

Statistics - Lecture One. Outline. Charlotte Wickham  1. Basic ideas about estimation Statistics - Lecture One Charlotte Wickham wickham@stat.berkeley.edu http://www.stat.berkeley.edu/~wickham/ Outline 1. Basic ideas about estimation 2. Method of Moments 3. Maximum Likelihood 4. Confidence

More information

Analysis of Type-II Progressively Hybrid Censored Data

Analysis of Type-II Progressively Hybrid Censored Data Analysis of Type-II Progressively Hybrid Censored Data Debasis Kundu & Avijit Joarder Abstract The mixture of Type-I and Type-II censoring schemes, called the hybrid censoring scheme is quite common in

More information

Lifetime prediction and confidence bounds in accelerated degradation testing for lognormal response distributions with an Arrhenius rate relationship

Lifetime prediction and confidence bounds in accelerated degradation testing for lognormal response distributions with an Arrhenius rate relationship Scholars' Mine Doctoral Dissertations Student Research & Creative Works Spring 01 Lifetime prediction and confidence bounds in accelerated degradation testing for lognormal response distributions with

More information

STAT 6350 Analysis of Lifetime Data. Probability Plotting

STAT 6350 Analysis of Lifetime Data. Probability Plotting STAT 6350 Analysis of Lifetime Data Probability Plotting Purpose of Probability Plots Probability plots are an important tool for analyzing data and have been particular popular in the analysis of life

More information

COPYRIGHTED MATERIAL CONTENTS. Preface Preface to the First Edition

COPYRIGHTED MATERIAL CONTENTS. Preface Preface to the First Edition Preface Preface to the First Edition xi xiii 1 Basic Probability Theory 1 1.1 Introduction 1 1.2 Sample Spaces and Events 3 1.3 The Axioms of Probability 7 1.4 Finite Sample Spaces and Combinatorics 15

More information

A Simulation Study on Confidence Interval Procedures of Some Mean Cumulative Function Estimators

A Simulation Study on Confidence Interval Procedures of Some Mean Cumulative Function Estimators Statistics Preprints Statistics -00 A Simulation Study on Confidence Interval Procedures of Some Mean Cumulative Function Estimators Jianying Zuo Iowa State University, jiyizu@iastate.edu William Q. Meeker

More information

Estimation of Quantiles

Estimation of Quantiles 9 Estimation of Quantiles The notion of quantiles was introduced in Section 3.2: recall that a quantile x α for an r.v. X is a constant such that P(X x α )=1 α. (9.1) In this chapter we examine quantiles

More information

Hybrid Censoring; An Introduction 2

Hybrid Censoring; An Introduction 2 Hybrid Censoring; An Introduction 2 Debasis Kundu Department of Mathematics & Statistics Indian Institute of Technology Kanpur 23-rd November, 2010 2 This is a joint work with N. Balakrishnan Debasis Kundu

More information

INFERENCE FOR A GAMMA STEP-STRESS MODEL UNDER CENSORING

INFERENCE FOR A GAMMA STEP-STRESS MODEL UNDER CENSORING INFERENCE FOR A GAMMA STEP-STRESS MODEL UNDER CENSORING INFERENCE FOR A GAMMA STEP-STRESS MODEL UNDER CENSORING BY LAILA ALKHALFAN, M.Sc. McMaster University a thesis submitted to the school of graduate

More information

Optimum Test Plan for 3-Step, Step-Stress Accelerated Life Tests

Optimum Test Plan for 3-Step, Step-Stress Accelerated Life Tests International Journal of Performability Engineering, Vol., No., January 24, pp.3-4. RAMS Consultants Printed in India Optimum Test Plan for 3-Step, Step-Stress Accelerated Life Tests N. CHANDRA *, MASHROOR

More information

Estimation of Parameters of the Weibull Distribution Based on Progressively Censored Data

Estimation of Parameters of the Weibull Distribution Based on Progressively Censored Data International Mathematical Forum, 2, 2007, no. 41, 2031-2043 Estimation of Parameters of the Weibull Distribution Based on Progressively Censored Data K. S. Sultan 1 Department of Statistics Operations

More information

The Relationship Between Confidence Intervals for Failure Probabilities and Life Time Quantiles

The Relationship Between Confidence Intervals for Failure Probabilities and Life Time Quantiles Statistics Preprints Statistics 2008 The Relationship Between Confidence Intervals for Failure Probabilities and Life Time Quantiles Yili Hong Iowa State University, yili_hong@hotmail.com William Q. Meeker

More information

INFORMATION APPROACH FOR CHANGE POINT DETECTION OF WEIBULL MODELS WITH APPLICATIONS. Tao Jiang. A Thesis

INFORMATION APPROACH FOR CHANGE POINT DETECTION OF WEIBULL MODELS WITH APPLICATIONS. Tao Jiang. A Thesis INFORMATION APPROACH FOR CHANGE POINT DETECTION OF WEIBULL MODELS WITH APPLICATIONS Tao Jiang A Thesis Submitted to the Graduate College of Bowling Green State University in partial fulfillment of the

More information

Simultaneous Prediction Intervals for the (Log)- Location-Scale Family of Distributions

Simultaneous Prediction Intervals for the (Log)- Location-Scale Family of Distributions Statistics Preprints Statistics 10-2014 Simultaneous Prediction Intervals for the (Log)- Location-Scale Family of Distributions Yimeng Xie Virginia Tech Yili Hong Virginia Tech Luis A. Escobar Louisiana

More information

A Tool for Evaluating Time-Varying-Stress Accelerated Life Test Plans with Log-Location- Scale Distributions

A Tool for Evaluating Time-Varying-Stress Accelerated Life Test Plans with Log-Location- Scale Distributions Statistics Preprints Statistics 6-2010 A Tool for Evaluating Time-Varying-Stress Accelerated Life Test Plans with Log-Location- Scale Distributions Yili Hong Virginia Tech Haiming Ma Iowa State University,

More information

STEP STRESS TESTS AND SOME EXACT INFERENTIAL RESULTS N. BALAKRISHNAN. McMaster University Hamilton, Ontario, Canada. p.

STEP STRESS TESTS AND SOME EXACT INFERENTIAL RESULTS N. BALAKRISHNAN. McMaster University Hamilton, Ontario, Canada. p. p. 1/6 STEP STRESS TESTS AND SOME EXACT INFERENTIAL RESULTS N. BALAKRISHNAN bala@mcmaster.ca McMaster University Hamilton, Ontario, Canada p. 2/6 In collaboration with Debasis Kundu, IIT, Kapur, India

More information

10 Introduction to Reliability

10 Introduction to Reliability 0 Introduction to Reliability 10 Introduction to Reliability The following notes are based on Volume 6: How to Analyze Reliability Data, by Wayne Nelson (1993), ASQC Press. When considering the reliability

More information

System Simulation Part II: Mathematical and Statistical Models Chapter 5: Statistical Models

System Simulation Part II: Mathematical and Statistical Models Chapter 5: Statistical Models System Simulation Part II: Mathematical and Statistical Models Chapter 5: Statistical Models Fatih Cavdur fatihcavdur@uludag.edu.tr March 20, 2012 Introduction Introduction The world of the model-builder

More information

The comparative studies on reliability for Rayleigh models

The comparative studies on reliability for Rayleigh models Journal of the Korean Data & Information Science Society 018, 9, 533 545 http://dx.doi.org/10.7465/jkdi.018.9..533 한국데이터정보과학회지 The comparative studies on reliability for Rayleigh models Ji Eun Oh 1 Joong

More information

Multistate Modeling and Applications

Multistate Modeling and Applications Multistate Modeling and Applications Yang Yang Department of Statistics University of Michigan, Ann Arbor IBM Research Graduate Student Workshop: Statistics for a Smarter Planet Yang Yang (UM, Ann Arbor)

More information

Parameter Estimation

Parameter Estimation Parameter Estimation Chapters 13-15 Stat 477 - Loss Models Chapters 13-15 (Stat 477) Parameter Estimation Brian Hartman - BYU 1 / 23 Methods for parameter estimation Methods for parameter estimation Methods

More information

Introduction to Reliability Theory (part 2)

Introduction to Reliability Theory (part 2) Introduction to Reliability Theory (part 2) Frank Coolen UTOPIAE Training School II, Durham University 3 July 2018 (UTOPIAE) Introduction to Reliability Theory 1 / 21 Outline Statistical issues Software

More information

n =10,220 observations. Smaller samples analyzed here to illustrate sample size effect.

n =10,220 observations. Smaller samples analyzed here to illustrate sample size effect. Chapter 7 Parametric Likelihood Fitting Concepts: Chapter 7 Parametric Likelihood Fitting Concepts: Objectives Show how to compute a likelihood for a parametric model using discrete data. Show how to compute

More information

Chapter 2 Inference on Mean Residual Life-Overview

Chapter 2 Inference on Mean Residual Life-Overview Chapter 2 Inference on Mean Residual Life-Overview Statistical inference based on the remaining lifetimes would be intuitively more appealing than the popular hazard function defined as the risk of immediate

More information

MAS3301 / MAS8311 Biostatistics Part II: Survival

MAS3301 / MAS8311 Biostatistics Part II: Survival MAS330 / MAS83 Biostatistics Part II: Survival M. Farrow School of Mathematics and Statistics Newcastle University Semester 2, 2009-0 8 Parametric models 8. Introduction In the last few sections (the KM

More information

Estimation and Confidence Intervals for Parameters of a Cumulative Damage Model

Estimation and Confidence Intervals for Parameters of a Cumulative Damage Model United States Department of Agriculture Forest Service Forest Products Laboratory Research Paper FPL-RP-484 Estimation and Confidence Intervals for Parameters of a Cumulative Damage Model Carol L. Link

More information

1 Degree distributions and data

1 Degree distributions and data 1 Degree distributions and data A great deal of effort is often spent trying to identify what functional form best describes the degree distribution of a network, particularly the upper tail of that distribution.

More information

Point and Interval Estimation for Gaussian Distribution, Based on Progressively Type-II Censored Samples

Point and Interval Estimation for Gaussian Distribution, Based on Progressively Type-II Censored Samples 90 IEEE TRANSACTIONS ON RELIABILITY, VOL. 52, NO. 1, MARCH 2003 Point and Interval Estimation for Gaussian Distribution, Based on Progressively Type-II Censored Samples N. Balakrishnan, N. Kannan, C. T.

More information

Some Theoretical Properties and Parameter Estimation for the Two-Sided Length Biased Inverse Gaussian Distribution

Some Theoretical Properties and Parameter Estimation for the Two-Sided Length Biased Inverse Gaussian Distribution Journal of Probability and Statistical Science 14(), 11-4, Aug 016 Some Theoretical Properties and Parameter Estimation for the Two-Sided Length Biased Inverse Gaussian Distribution Teerawat Simmachan

More information

Mathematical statistics

Mathematical statistics October 1 st, 2018 Lecture 11: Sufficient statistic Where are we? Week 1 Week 2 Week 4 Week 7 Week 10 Week 14 Probability reviews Chapter 6: Statistics and Sampling Distributions Chapter 7: Point Estimation

More information

INVERTED KUMARASWAMY DISTRIBUTION: PROPERTIES AND ESTIMATION

INVERTED KUMARASWAMY DISTRIBUTION: PROPERTIES AND ESTIMATION Pak. J. Statist. 2017 Vol. 33(1), 37-61 INVERTED KUMARASWAMY DISTRIBUTION: PROPERTIES AND ESTIMATION A. M. Abd AL-Fattah, A.A. EL-Helbawy G.R. AL-Dayian Statistics Department, Faculty of Commerce, AL-Azhar

More information

Optimum Hybrid Censoring Scheme using Cost Function Approach

Optimum Hybrid Censoring Scheme using Cost Function Approach Optimum Hybrid Censoring Scheme using Cost Function Approach Ritwik Bhattacharya 1, Biswabrata Pradhan 1, Anup Dewanji 2 1 SQC and OR Unit, Indian Statistical Institute, 203, B. T. Road, Kolkata, PIN-

More information

Statistical Methods for Reliability Data from Designed Experiments

Statistical Methods for Reliability Data from Designed Experiments Statistical Methods for Reliability Data from Designed Experiments Laura J. Freeman Dissertation submitted to the Faculty of the Virginia Polytechnic Institute and State University in partial fulfillment

More information

Practice Problems Section Problems

Practice Problems Section Problems Practice Problems Section 4-4-3 4-4 4-5 4-6 4-7 4-8 4-10 Supplemental Problems 4-1 to 4-9 4-13, 14, 15, 17, 19, 0 4-3, 34, 36, 38 4-47, 49, 5, 54, 55 4-59, 60, 63 4-66, 68, 69, 70, 74 4-79, 81, 84 4-85,

More information

Reliability analysis under constant-stress partially accelerated life tests using hybrid censored data from Weibull distribution

Reliability analysis under constant-stress partially accelerated life tests using hybrid censored data from Weibull distribution Hacettepe Journal of Mathematics and Statistics Volume 45 (1) (2016), 181 193 Reliability analysis under constant-stress partially accelerated life tests using hybrid censored data from Weibull distribution

More information

Likelihood Construction, Inference for Parametric Survival Distributions

Likelihood Construction, Inference for Parametric Survival Distributions Week 1 Likelihood Construction, Inference for Parametric Survival Distributions In this section we obtain the likelihood function for noninformatively rightcensored survival data and indicate how to make

More information

Mathematical statistics

Mathematical statistics October 4 th, 2018 Lecture 12: Information Where are we? Week 1 Week 2 Week 4 Week 7 Week 10 Week 14 Probability reviews Chapter 6: Statistics and Sampling Distributions Chapter 7: Point Estimation Chapter

More information

Estimation for inverse Gaussian Distribution Under First-failure Progressive Hybird Censored Samples

Estimation for inverse Gaussian Distribution Under First-failure Progressive Hybird Censored Samples Filomat 31:18 (217), 5743 5752 https://doi.org/1.2298/fil1718743j Published by Faculty of Sciences and Mathematics, University of Niš, Serbia Available at: http://www.pmf.ni.ac.rs/filomat Estimation for

More information

Statistics 3858 : Maximum Likelihood Estimators

Statistics 3858 : Maximum Likelihood Estimators Statistics 3858 : Maximum Likelihood Estimators 1 Method of Maximum Likelihood In this method we construct the so called likelihood function, that is L(θ) = L(θ; X 1, X 2,..., X n ) = f n (X 1, X 2,...,

More information

3 Continuous Random Variables

3 Continuous Random Variables Jinguo Lian Math437 Notes January 15, 016 3 Continuous Random Variables Remember that discrete random variables can take only a countable number of possible values. On the other hand, a continuous random

More information

Exponentiated Rayleigh Distribution: A Bayes Study Using MCMC Approach Based on Unified Hybrid Censored Data

Exponentiated Rayleigh Distribution: A Bayes Study Using MCMC Approach Based on Unified Hybrid Censored Data Exponentiated Rayleigh Distribution: A Bayes Study Using MCMC Approach Based on Unified Hybrid Censored Data ABSTRACT M. G. M. Ghazal 1, H. M. Hasaballah 2 1 Mathematics Department, Faculty of Science,

More information

Chapter 17. Failure-Time Regression Analysis. William Q. Meeker and Luis A. Escobar Iowa State University and Louisiana State University

Chapter 17. Failure-Time Regression Analysis. William Q. Meeker and Luis A. Escobar Iowa State University and Louisiana State University Chapter 17 Failure-Time Regression Analysis William Q. Meeker and Luis A. Escobar Iowa State University and Louisiana State University Copyright 1998-2008 W. Q. Meeker and L. A. Escobar. Based on the authors

More information

A Very Brief Summary of Statistical Inference, and Examples

A Very Brief Summary of Statistical Inference, and Examples A Very Brief Summary of Statistical Inference, and Examples Trinity Term 2009 Prof. Gesine Reinert Our standard situation is that we have data x = x 1, x 2,..., x n, which we view as realisations of random

More information

Practice Exam 1. (A) (B) (C) (D) (E) You are given the following data on loss sizes:

Practice Exam 1. (A) (B) (C) (D) (E) You are given the following data on loss sizes: Practice Exam 1 1. Losses for an insurance coverage have the following cumulative distribution function: F(0) = 0 F(1,000) = 0.2 F(5,000) = 0.4 F(10,000) = 0.9 F(100,000) = 1 with linear interpolation

More information

Bayesian Life Test Planning for the Weibull Distribution with Given Shape Parameter

Bayesian Life Test Planning for the Weibull Distribution with Given Shape Parameter Statistics Preprints Statistics 10-8-2002 Bayesian Life Test Planning for the Weibull Distribution with Given Shape Parameter Yao Zhang Iowa State University William Q. Meeker Iowa State University, wqmeeker@iastate.edu

More information

Distribution Fitting (Censored Data)

Distribution Fitting (Censored Data) Distribution Fitting (Censored Data) Summary... 1 Data Input... 2 Analysis Summary... 3 Analysis Options... 4 Goodness-of-Fit Tests... 6 Frequency Histogram... 8 Comparison of Alternative Distributions...

More information

Statistical Inference Using Progressively Type-II Censored Data with Random Scheme

Statistical Inference Using Progressively Type-II Censored Data with Random Scheme International Mathematical Forum, 3, 28, no. 35, 1713-1725 Statistical Inference Using Progressively Type-II Censored Data with Random Scheme Ammar M. Sarhan 1 and A. Abuammoh Department of Statistics

More information

Bootstrap Confidence Intervals

Bootstrap Confidence Intervals Bootstrap Confidence Intervals Patrick Breheny September 18 Patrick Breheny STA 621: Nonparametric Statistics 1/22 Introduction Bootstrap confidence intervals So far, we have discussed the idea behind

More information

Inference on reliability in two-parameter exponential stress strength model

Inference on reliability in two-parameter exponential stress strength model Metrika DOI 10.1007/s00184-006-0074-7 Inference on reliability in two-parameter exponential stress strength model K. Krishnamoorthy Shubhabrata Mukherjee Huizhen Guo Received: 19 January 2005 Springer-Verlag

More information

Math 494: Mathematical Statistics

Math 494: Mathematical Statistics Math 494: Mathematical Statistics Instructor: Jimin Ding jmding@wustl.edu Department of Mathematics Washington University in St. Louis Class materials are available on course website (www.math.wustl.edu/

More information

Analysis of Time-to-Event Data: Chapter 4 - Parametric regression models

Analysis of Time-to-Event Data: Chapter 4 - Parametric regression models Analysis of Time-to-Event Data: Chapter 4 - Parametric regression models Steffen Unkel Department of Medical Statistics University Medical Center Göttingen, Germany Winter term 2018/19 1/25 Right censored

More information

Estimation for generalized half logistic distribution based on records

Estimation for generalized half logistic distribution based on records Journal of the Korean Data & Information Science Society 202, 236, 249 257 http://dx.doi.org/0.7465/jkdi.202.23.6.249 한국데이터정보과학회지 Estimation for generalized half logistic distribution based on records

More information

Bivariate Degradation Modeling Based on Gamma Process

Bivariate Degradation Modeling Based on Gamma Process Bivariate Degradation Modeling Based on Gamma Process Jinglun Zhou Zhengqiang Pan Member IAENG and Quan Sun Abstract Many highly reliable products have two or more performance characteristics (PCs). The

More information

Inferences about Parameters of Trivariate Normal Distribution with Missing Data

Inferences about Parameters of Trivariate Normal Distribution with Missing Data Florida International University FIU Digital Commons FIU Electronic Theses and Dissertations University Graduate School 7-5-3 Inferences about Parameters of Trivariate Normal Distribution with Missing

More information

Better Bootstrap Confidence Intervals

Better Bootstrap Confidence Intervals by Bradley Efron University of Washington, Department of Statistics April 12, 2012 An example Suppose we wish to make inference on some parameter θ T (F ) (e.g. θ = E F X ), based on data We might suppose

More information

Hybrid Censoring Scheme: An Introduction

Hybrid Censoring Scheme: An Introduction Department of Mathematics & Statistics Indian Institute of Technology Kanpur August 19, 2014 Outline 1 2 3 4 5 Outline 1 2 3 4 5 What is? Lifetime data analysis is used to analyze data in which the time

More information

f(x θ)dx with respect to θ. Assuming certain smoothness conditions concern differentiating under the integral the integral sign, we first obtain

f(x θ)dx with respect to θ. Assuming certain smoothness conditions concern differentiating under the integral the integral sign, we first obtain 0.1. INTRODUCTION 1 0.1 Introduction R. A. Fisher, a pioneer in the development of mathematical statistics, introduced a measure of the amount of information contained in an observaton from f(x θ). Fisher

More information

Unit 20: Planning Accelerated Life Tests

Unit 20: Planning Accelerated Life Tests Unit 20: Planning Accelerated Life Tests Ramón V. León Notes largely based on Statistical Methods for Reliability Data by W.Q. Meeker and L. A. Escobar, Wiley, 1998 and on their class notes. 11/13/2004

More information

ST745: Survival Analysis: Nonparametric methods

ST745: Survival Analysis: Nonparametric methods ST745: Survival Analysis: Nonparametric methods Eric B. Laber Department of Statistics, North Carolina State University February 5, 2015 The KM estimator is used ubiquitously in medical studies to estimate

More information

STATS 200: Introduction to Statistical Inference. Lecture 29: Course review

STATS 200: Introduction to Statistical Inference. Lecture 29: Course review STATS 200: Introduction to Statistical Inference Lecture 29: Course review Course review We started in Lecture 1 with a fundamental assumption: Data is a realization of a random process. The goal throughout

More information

Time-varying failure rate for system reliability analysis in large-scale railway risk assessment simulation

Time-varying failure rate for system reliability analysis in large-scale railway risk assessment simulation Time-varying failure rate for system reliability analysis in large-scale railway risk assessment simulation H. Zhang, E. Cutright & T. Giras Center of Rail Safety-Critical Excellence, University of Virginia,

More information

Parameter Estimation of Power Lomax Distribution Based on Type-II Progressively Hybrid Censoring Scheme

Parameter Estimation of Power Lomax Distribution Based on Type-II Progressively Hybrid Censoring Scheme Applied Mathematical Sciences, Vol. 12, 2018, no. 18, 879-891 HIKARI Ltd, www.m-hikari.com https://doi.org/10.12988/ams.2018.8691 Parameter Estimation of Power Lomax Distribution Based on Type-II Progressively

More information

1. Point Estimators, Review

1. Point Estimators, Review AMS571 Prof. Wei Zhu 1. Point Estimators, Review Example 1. Let be a random sample from. Please find a good point estimator for Solutions. There are the typical estimators for and. Both are unbiased estimators.

More information

STAT 512 sp 2018 Summary Sheet

STAT 512 sp 2018 Summary Sheet STAT 5 sp 08 Summary Sheet Karl B. Gregory Spring 08. Transformations of a random variable Let X be a rv with support X and let g be a function mapping X to Y with inverse mapping g (A = {x X : g(x A}

More information

Approximation of Survival Function by Taylor Series for General Partly Interval Censored Data

Approximation of Survival Function by Taylor Series for General Partly Interval Censored Data Malaysian Journal of Mathematical Sciences 11(3): 33 315 (217) MALAYSIAN JOURNAL OF MATHEMATICAL SCIENCES Journal homepage: http://einspem.upm.edu.my/journal Approximation of Survival Function by Taylor

More information

Bayesian Analysis of Simple Step-stress Model under Weibull Lifetimes

Bayesian Analysis of Simple Step-stress Model under Weibull Lifetimes Bayesian Analysis of Simple Step-stress Model under Weibull Lifetimes A. Ganguly 1, D. Kundu 2,3, S. Mitra 2 Abstract Step-stress model is becoming quite popular in recent times for analyzing lifetime

More information

ST495: Survival Analysis: Hypothesis testing and confidence intervals

ST495: Survival Analysis: Hypothesis testing and confidence intervals ST495: Survival Analysis: Hypothesis testing and confidence intervals Eric B. Laber Department of Statistics, North Carolina State University April 3, 2014 I remember that one fateful day when Coach took

More information

Confidence Intervals in Ridge Regression using Jackknife and Bootstrap Methods

Confidence Intervals in Ridge Regression using Jackknife and Bootstrap Methods Chapter 4 Confidence Intervals in Ridge Regression using Jackknife and Bootstrap Methods 4.1 Introduction It is now explicable that ridge regression estimator (here we take ordinary ridge estimator (ORE)

More information

Interval Estimation for Parameters of a Bivariate Time Varying Covariate Model

Interval Estimation for Parameters of a Bivariate Time Varying Covariate Model Pertanika J. Sci. & Technol. 17 (2): 313 323 (2009) ISSN: 0128-7680 Universiti Putra Malaysia Press Interval Estimation for Parameters of a Bivariate Time Varying Covariate Model Jayanthi Arasan Department

More information

Bayesian Inference and Life Testing Plan for the Weibull Distribution in Presence of Progressive Censoring

Bayesian Inference and Life Testing Plan for the Weibull Distribution in Presence of Progressive Censoring Bayesian Inference and Life Testing Plan for the Weibull Distribution in Presence of Progressive Censoring Debasis KUNDU Department of Mathematics and Statistics Indian Institute of Technology Kanpur Pin

More information

CHAPTER 3 ANALYSIS OF RELIABILITY AND PROBABILITY MEASURES

CHAPTER 3 ANALYSIS OF RELIABILITY AND PROBABILITY MEASURES 27 CHAPTER 3 ANALYSIS OF RELIABILITY AND PROBABILITY MEASURES 3.1 INTRODUCTION The express purpose of this research is to assimilate reliability and its associated probabilistic variables into the Unit

More information

Estimation MLE-Pandemic data MLE-Financial crisis data Evaluating estimators. Estimation. September 24, STAT 151 Class 6 Slide 1

Estimation MLE-Pandemic data MLE-Financial crisis data Evaluating estimators. Estimation. September 24, STAT 151 Class 6 Slide 1 Estimation September 24, 2018 STAT 151 Class 6 Slide 1 Pandemic data Treatment outcome, X, from n = 100 patients in a pandemic: 1 = recovered and 0 = not recovered 1 1 1 0 0 0 1 1 1 0 0 1 0 1 0 0 1 1 1

More information

OPTIMUM DESIGN ON STEP-STRESS LIFE TESTING

OPTIMUM DESIGN ON STEP-STRESS LIFE TESTING Libraries Conference on Applied Statistics in Agriculture 1998-10th Annual Conference Proceedings OPTIMUM DESIGN ON STEP-STRESS LIFE TESTING C. Xiong Follow this and additional works at: http://newprairiepress.org/agstatconference

More information

An Analysis of Record Statistics based on an Exponentiated Gumbel Model

An Analysis of Record Statistics based on an Exponentiated Gumbel Model Communications for Statistical Applications and Methods 2013, Vol. 20, No. 5, 405 416 DOI: http://dx.doi.org/10.5351/csam.2013.20.5.405 An Analysis of Record Statistics based on an Exponentiated Gumbel

More information

Computational Statistics and Data Analysis. Estimation for the three-parameter lognormal distribution based on progressively censored data

Computational Statistics and Data Analysis. Estimation for the three-parameter lognormal distribution based on progressively censored data Computational Statistics and Data Analysis 53 (9) 358 359 Contents lists available at ScienceDirect Computational Statistics and Data Analysis journal homepage: www.elsevier.com/locate/csda stimation for

More information

Accelerated Destructive Degradation Test Planning

Accelerated Destructive Degradation Test Planning Accelerated Destructive Degradation Test Planning Ying Shi Dept. of Statistics Iowa State University Ames, IA 50011 yshi@iastate.edu Luis A. Escobar Dept. of Experimental Statistics Louisiana State University

More information

Stat 5101 Lecture Notes

Stat 5101 Lecture Notes Stat 5101 Lecture Notes Charles J. Geyer Copyright 1998, 1999, 2000, 2001 by Charles J. Geyer May 7, 2001 ii Stat 5101 (Geyer) Course Notes Contents 1 Random Variables and Change of Variables 1 1.1 Random

More information

Statistical Estimation

Statistical Estimation Statistical Estimation Use data and a model. The plug-in estimators are based on the simple principle of applying the defining functional to the ECDF. Other methods of estimation: minimize residuals from

More information

Statistical Prediction Based on Censored Life Data. Luis A. Escobar Department of Experimental Statistics Louisiana State University.

Statistical Prediction Based on Censored Life Data. Luis A. Escobar Department of Experimental Statistics Louisiana State University. Statistical Prediction Based on Censored Life Data Overview Luis A. Escobar Department of Experimental Statistics Louisiana State University and William Q. Meeker Department of Statistics Iowa State University

More information

Slides 8: Statistical Models in Simulation

Slides 8: Statistical Models in Simulation Slides 8: Statistical Models in Simulation Purpose and Overview The world the model-builder sees is probabilistic rather than deterministic: Some statistical model might well describe the variations. An

More information

Estimation for Mean and Standard Deviation of Normal Distribution under Type II Censoring

Estimation for Mean and Standard Deviation of Normal Distribution under Type II Censoring Communications for Statistical Applications and Methods 2014, Vol. 21, No. 6, 529 538 DOI: http://dx.doi.org/10.5351/csam.2014.21.6.529 Print ISSN 2287-7843 / Online ISSN 2383-4757 Estimation for Mean

More information

Analysis of Gamma and Weibull Lifetime Data under a General Censoring Scheme and in the presence of Covariates

Analysis of Gamma and Weibull Lifetime Data under a General Censoring Scheme and in the presence of Covariates Communications in Statistics - Theory and Methods ISSN: 0361-0926 (Print) 1532-415X (Online) Journal homepage: http://www.tandfonline.com/loi/lsta20 Analysis of Gamma and Weibull Lifetime Data under a

More information

Semiparametric Regression

Semiparametric Regression Semiparametric Regression Patrick Breheny October 22 Patrick Breheny Survival Data Analysis (BIOS 7210) 1/23 Introduction Over the past few weeks, we ve introduced a variety of regression models under

More information

Statistics: Learning models from data

Statistics: Learning models from data DS-GA 1002 Lecture notes 5 October 19, 2015 Statistics: Learning models from data Learning models from data that are assumed to be generated probabilistically from a certain unknown distribution is a crucial

More information

Unit 10: Planning Life Tests

Unit 10: Planning Life Tests Unit 10: Planning Life Tests Ramón V. León Notes largely based on Statistical Methods for Reliability Data by W.Q. Meeker and L. A. Escobar, Wiley, 1998 and on their class notes. 11/2/2004 Unit 10 - Stat

More information

Robustness and Distribution Assumptions

Robustness and Distribution Assumptions Chapter 1 Robustness and Distribution Assumptions 1.1 Introduction In statistics, one often works with model assumptions, i.e., one assumes that data follow a certain model. Then one makes use of methodology

More information

Fixed Point Iteration for Estimating The Parameters of Extreme Value Distributions

Fixed Point Iteration for Estimating The Parameters of Extreme Value Distributions Fixed Point Iteration for Estimating The Parameters of Extreme Value Distributions arxiv:0902.07v [stat.co] Feb 2009 Tewfik Kernane and Zohrh A. Raizah 2 Department of Mathematics, Faculty of Sciences

More information

LOGISTIC REGRESSION Joseph M. Hilbe

LOGISTIC REGRESSION Joseph M. Hilbe LOGISTIC REGRESSION Joseph M. Hilbe Arizona State University Logistic regression is the most common method used to model binary response data. When the response is binary, it typically takes the form of

More information

Lecture 3 September 1

Lecture 3 September 1 STAT 383C: Statistical Modeling I Fall 2016 Lecture 3 September 1 Lecturer: Purnamrita Sarkar Scribe: Giorgio Paulon, Carlos Zanini Disclaimer: These scribe notes have been slightly proofread and may have

More information

SOLUTION FOR HOMEWORK 8, STAT 4372

SOLUTION FOR HOMEWORK 8, STAT 4372 SOLUTION FOR HOMEWORK 8, STAT 4372 Welcome to your 8th homework. Here you have an opportunity to solve classical estimation problems which are the must to solve on the exam due to their simplicity. 1.

More information

Mathematical statistics

Mathematical statistics October 18 th, 2018 Lecture 16: Midterm review Countdown to mid-term exam: 7 days Week 1 Chapter 1: Probability review Week 2 Week 4 Week 7 Chapter 6: Statistics Chapter 7: Point Estimation Chapter 8:

More information

Structural Reliability

Structural Reliability Structural Reliability Thuong Van DANG May 28, 2018 1 / 41 2 / 41 Introduction to Structural Reliability Concept of Limit State and Reliability Review of Probability Theory First Order Second Moment Method

More information

Estimation in an Exponentiated Half Logistic Distribution under Progressively Type-II Censoring

Estimation in an Exponentiated Half Logistic Distribution under Progressively Type-II Censoring Communications of the Korean Statistical Society 2011, Vol. 18, No. 5, 657 666 DOI: http://dx.doi.org/10.5351/ckss.2011.18.5.657 Estimation in an Exponentiated Half Logistic Distribution under Progressively

More information

Different methods of estimation for generalized inverse Lindley distribution

Different methods of estimation for generalized inverse Lindley distribution Different methods of estimation for generalized inverse Lindley distribution Arbër Qoshja & Fatmir Hoxha Department of Applied Mathematics, Faculty of Natural Science, University of Tirana, Albania,, e-mail:

More information

Parameters Estimation for a Linear Exponential Distribution Based on Grouped Data

Parameters Estimation for a Linear Exponential Distribution Based on Grouped Data International Mathematical Forum, 3, 2008, no. 33, 1643-1654 Parameters Estimation for a Linear Exponential Distribution Based on Grouped Data A. Al-khedhairi Department of Statistics and O.R. Faculty

More information