A study of two estimation approaches for parameters of Weibull distribution based on WPP
Introduction
The Weibull distribution is widely used in life testing and reliability studies. Some modifications have been developed in recent years, which greatly enlarged the applications. For some recent references, see e.g. [1], [2], [3], [4].
The cumulative distribution function (CDF) of the two-parameter Weibull distribution iswhere t is the time to failure, α is the scale parameter and β is the shape parameter. The shape parameter characterizes the failure rate trend [5] and thus can indicate some failure modes (initial, random and wear-out) [6].
Many methods have been proposed for estimating α and β, such as the method of moments and the maximum likelihood estimation (MLE). Robust estimation methods and bias correction methods are of recent interests; see, e.g., [7], [8]. For simplicity, a common practice is to plot the failure time t vs. the estimated F(t) on a Weibull probability paper, and then fit a straight line to the data points. β can be estimated by the slope of the straight line and α by the intercept and slope. This method is preferred by engineers because of its computation simplicity and the graphical presentation [9]. The Weibull probability plot (WPP), in addition to providing simple straight line fitting for parameter estimation, serves other purposes, including evaluating whether the data are Weibull distributed and identifying outliers [10].
After data points are placed on the Weibull paper, the least-squares (LS) method is usually used to fit a straight line instead of by eye. The straight line can be fit by minimizing the sum of squares of the vertical residuals, which is called LS Y on X method, or it can be fit by minimizing the sum of squares of the horizontal residuals, which is called LS X on Y method. This paper compares these two methods for estimating the Weibull parameters in cases of both complete and censored data. The motivation of our study is that we noticed LS X on Y was adopted in the early literature, see e.g., [11], [12], [13], but is replaced by LS Y on X in the recent literature with a few exceptions, see, e.g., [14], [15].
To the best of our knowledge, there is no intensive comparison of the two methods existing in the literature, especially for censored data. In this paper, the two methods are compared in terms of both properties of the parameter estimators and model statistics of the two linear regression models. Both complete and multiply censored samples are examined. Suggestions of when to use which method are given based on the study.
Section snippets
Estimators of LS Y on X and LS X on Y
Suppose that t1,t2,…,tr(r⩽n) is a random sample from a life testing experiment with n items, and r<n refers to a censored sample. t(1),t(2),…,t(r) are the order statistics. Assuming the underlying distribution of the failure times is a 2-parameter Weibull distribution with unknown parameters α and β. Let a and b denote the estimators of α and β.
By taking logarithm twice, Eq. (1) can be linearized, i.e.,For the sample, Eq. (2) becomeswhere i
Properties of the LS estimators
Properties of an estimator include unbiasedness, consistency, efficiency, mean square error (MSE), etc. For a simple linear regression model of the form Y=A+BX+e, where e is the error term or residual, it is well known that the LS Y on X estimators of A and B are unbiased and efficient if the assumptions that e is approximately normally distributed with mean zero and a constant variance, and any pairs of ei, ej are uncorrelated, are satisfied. At the same time, it is clear that the LS X on Y
Monte Carlo study of the two methods
Monte Carlo simulation has been carried out to compare the performance of LS Y on X and LS X on Y on parameter estimation for complete and multiply censored samples of different sizes. Taking advantage of the first pivotal function of the LS estimators, we only need to generate random samples from the normalized Weibull distribution and compare b1,1 obtained by different methods to make a conclusion on which method is better for estimating β. However, a/α is not a pivotal function and is
Discussion
In this paper, two LS regression methods, LS Y on X and LS X on Y, are compared for estimating the two parameters for the Weibull distribution. The properties of their estimators are compared via the Monte Carlo method. With intensive simulations, some conclusions are made which can be used to give some suggestions on when to use which method. The methods are also compared in terms of model statistics. We find which model is better can be determined by the value of β. When β<1, LS Y on X will
Acknowledgment
The authors would like to thank a referee for the useful comments on an earlier version of the paper.
References (26)
- et al.
On Changing Points of Mean Residual Life and Failure Rate Function for Some Generalized Weibull Distributions
Reliab Eng Syst Saf
(2004) On the Moments of the Modified Weibull Distribution
Reliab Eng Syst Saf
(2005)- et al.
Weibull Model Selection for Reliability Modelling
Reliab Eng Syst Saf
(2004) Parameter Estimation for a Modified Weibull Distribution, for Progressively Type-II Censored Samples
IEEE Trans Reliab
(2005)- et al.
Efficient Estimation of the Weibull Shape Parameter Based on a Modified Profile Likelihood
J Stat Comput Simulation
(2002) - et al.
Reliability Tests for Weibull Distribution with Varying Shape-parameter, based on Complete Data
IEEE Trans on Reliab
(2002) - et al.
Robust Weighted Likelihood Estimation of Exponential Parameters
IEEE Trans Reliab
(2005) - et al.
Bias Correction for Mean Time to Failure and p-quantiles in a Weibull Distribution by Bootstrap Procedure
Communi Stat-Simulation Comput
(2005) Refined Rank Regression Method with Censors
Qual Reliab Eng Int
(2004)Graphical Methods for Plotting and Evaluating Weibull Distributed Data
IEEE Trans Dielectr Electr Insul
(1994)
Fatigue testing and analysis of results, published for the Advisory Group for Aeronautical Research and Development, North Atlantic Treaty Organization
The Moments of Log-Weibull Order Statistics
Technometrics
Methods for statistical analysis of reliability and life data
Cited by (74)
Evaluation of roller compacted concrete for its application as high traffic resisting pavements with fatigue analysis
2023, Construction and Building MaterialsA new extension of the modified Weibull distribution with applications for engineering data
2023, Probabilistic Engineering MechanicsReliability estimation for two-parameter Weibull distribution under block censoring
2020, Reliability Engineering and System SafetyCitation Excerpt :In the remainder of the paper, we use Weibull distribution instead of two-parameter Weibull distribution for the sake of brevity. Extensive work has been done on the Weibull distribution using classical and Bayes approaches, see for example [18–23]. The censoring is quite common and popular for reducing test time and saving test cost in many life testing and reliability analysis.
Reliability analysis for Weibull distribution with homogeneous heavily censored data based on Bayesian and least-squares methods
2020, Applied Mathematical ModellingPerformance Evaluation of Ultra-high Performance Concrete (UHPC) and Ultra-high Performance Fibre Reinforced Concrete (UHPFRC) in Pavement Applications
2024, Arabian Journal for Science and EngineeringResearch on Data-Driven Approaches for Life Prediction of YBCO Tapes Under Overcurrent
2023, IEEE Transactions on Applied Superconductivity