摘要:

Following Gehan (1965)and Breslow(1970), a generalization of Steel's(1959)test for comparing several treatments with acontrol in a one-way layout when observations are subject to the same pattern of random right-censorship is proposed. The proposed test is constructed mainly for testing against simple-tree alternatives. However, based on the test, a multiple testing procedure for deciding which treatments(if any)are better than the control is suggested. The relative level and powerperformances of the proposed testing procedure and the ones suggested respectively by Magel(1988)and Chakraborti(1990) areexamined in a Monts\e Carlo study

ISSN:0361-0918    

DOI:10.1080/03610919408813152

出版商:Marcel Dekker, Inc.    

年代:1994

数据来源: Taylor

摘要:

Holmes and Mergen(1992)proposed the use of parabolic control limits for exponentially weighted moving average(EWMA)controlcharts.It is shown in an example that the usual control limits for the EWMA control chart perform much better than theparabolic control limits and are computationally more efficient

ISSN:0361-0918    

DOI:10.1080/03610919408813153

出版商:Marcel Dekker, Inc.    

年代:1994

数据来源: Taylor

摘要:

Simplified formulae are given for computing type I and type II error probabilities in a two stress level accelerated lif test plan developed by Yum and Kim(1990)Using the new formulae, some acceptance sampling plans for the mean value of anexponential distribution, based on type II censored data, are obtained. The criterion for choosing the censoring numbers, r1and rh,corresponding to a pair of higher-than-normal-use stress levels, is to minimize the sum, r1+rh.An approximate procedure is also included

ISSN:0361-0918    

DOI:10.1080/03610919408813154

出版商:Marcel Dekker, Inc.    

年代:1994

数据来源: Taylor

摘要:

In applications that utilize a linear regression model with nested error structure, interest often focuses on inferences concerning the regression coefficient. This article derives exact and approximate confidence intervals on the regression coefficient in the simple linear regression model with a balanced one-fold nested error structure. Several methods are considered for constructing the confidence intervals. The methods are compared using computer simulation and recommendations are provided for selecting an appropriate method

ISSN:0361-0918    

DOI:10.1080/03610919408813155

出版商:Marcel Dekker, Inc.    

年代:1994

数据来源: Taylor

摘要:

Maximum likelihood estimates (MLEs) for logistic regression coefficients are known to be biased in finite samples and consequently may produce misleading inferences. Bias adjusted estimates can be calculated using the first-order asymptotic bias derived from a Taylor series expansion of the log likelihood. Jackknifing can also be used to obtain bias corrected estimates, but the approach is computationally intensive, requiring an additional series of iterations (steps) for each observation in the dataset.Although the one-step jackknife has been shown to be useful in logistic regression diagnostics and i the estimation of classification error rates, it does not effectively reduce bias. The two-step jackknife, however, can reduce computation in moderate-sized samples, provide estimates of dispersion and classification error, and appears to be effective in bias reduction. Another alternative, a two-step closed-form approximation, is found to be similar to the Taylo series method in certain circumstances. Monte Carlo simulations indicate that all the procedures, but particularly the multi-step jackknife, may tend to over-correct in very small samples. Comparison of the various bias correction proceduresin an example from the medical literature illustrates that bias correction can have a considerable impact on inference

ISSN:0361-0918    

DOI:10.1080/03610919408813156

出版商:Marcel Dekker, Inc.    

年代:1994

数据来源: Taylor

摘要:

The commonly used age dependent preventive maintenance policies, for a repairable system, are based on the minimal repair model. But proportional hazards model (Cox, 1972; Lawless, 1987) is another reasonable model which will account for the possibility of additional damage at failures. Therefore, a test designed to discriminate between these two models will be useful. Such a test is developed and presented here. This test will be applicable when a large number of system copies, experiencing only a few failures each, are available for analysis. Initially, the score test is proposed under the assumption of known baseline rate of occurrence of failures of the system. Later this assumption is relaxed. The problem is then formulated in the framework of marked point processes. Tests based on partial likelihood are constructed. Further, a martingale estimator of the integrated baseline intensity is suggested. Performance of the tests is investigated by computing empirical powers for samples of sizes 50, 75 and 100

ISSN:0361-0918    

DOI:10.1080/03610919408813157

出版商:Marcel Dekker, Inc.    

年代:1994

数据来源: Taylor

摘要:

In the typical repeated measures analysis, the subject effect is assumed to be random and the error terms random and uncorrelated. (One could assume that the subject effect is fixed while the errors are correlated.) In many statistical packages, unless the user specifically requests that otherMean Squares be used for tests, the MSE is used by default. A simulation study investigates the magnitude of the errors that result if the inappropriate tests are performed. It is found that both the Type I and Type II errors can be greatly affected

ISSN:0361-0918    

DOI:10.1080/03610919408813158

出版商:Marcel Dekker, Inc.    

年代:1994

数据来源: Taylor

摘要:

The empirical performance of linear mean estimators satisfying some optimality conditions was studied and found to excel that of the arithmetic mean estimator particularly for processes generated by k-th differences in simulated data sets and empirical time series.

ISSN:0361-0918    

DOI:10.1080/03610919408813159

出版商:Marcel Dekker, Inc.    

年代:1994

数据来源: Taylor

摘要:

Two-factor experiments in which both factors are ordinal are considered. If it is believed apriorithat the mean response is nondecreasing in each factor with the other held fixed, then one may test for a treatment effect by testing homogeneitywith the appropriate ordered alternative. However, if this ordering were in question, one might test it as the null hypothesis. The likelihood ratio tests for both of these situations have been developed, but the level probabilities needed to implement these tests have only been determined in a few special cases. Monte Carlo estimates of the level probabilities are given here for balanced designs with no more than nine levels on each factor. These level probabilities are also useful when testing these hypotheses in one-parameter exponential families and when testing hypotheses involving a bivariate stochastic ordering. Monte Carlo estimates of the constants which are needed to estimate certain powers in these two-factor experiments are also given.

ISSN:0361-0918    

DOI:10.1080/03610919408813160

出版商:Marcel Dekker, Inc.    

年代:1994

数据来源: Taylor

摘要:

Ulrich's (1984) proposal for simulating unit vectors from the von Mises-Fisher distribution is discussed and an alternative specification of the algorithm is given. Then we describe an application to the simulation of the von Mises-Fisher matrix distribution for rotations ofR3or,equivalently, the Bingham distribution on the unit sphere inR4The same idea also leadsto a procedure for simulating the Bingham distribution on the unit sphere inRqwhen q > 4.

ISSN:0361-0918    

DOI:10.1080/03610919408813161

出版商:Marcel Dekker, Inc.    

年代:1994

数据来源: Taylor