摘要:

A Monte-Carlo simulation study is performed to compare the robustness to the assumption of normality of the discriminant function (DF) and conditional maximum likelihood (ML) methods of estimating the logistic regression coefficient (beta) for the simplified case of one independent variable. The unconditional probability of an event occurring, sample size, and beta are varied for each of four distributions - the normal, exponential, Bemouilli, and Poisson. A study of the bias of the estimate of beta, the standard deviation of the regression coefficient estimates and the significance level of the test of the coefficients suggests that, as expected, for very nonnormal distributions and moderate to large sample size, the ML estimates which do not require the normality assumption are preferred; however, for small sample sizes and highly significant beta, alternate methods to both procedures should be sought.

ISSN:0361-0918    

DOI:10.1080/03610918808812647

出版商:Marcel Dekker, Inc.    

年代:1988

数据来源: Taylor

摘要:

In this paper, we first obtain three generalized forms of the standard logistic distribution. These are derived by compounding an extreme-value distribution of the doubly exponential type, a reduced log-Weibull distribution and an exponential-gamma distribution, respectively, with a gamma distribution. The usefulness of these three types of generalized logistic distributions in some estimation problems, and also in some robustness studies, is pointed out. Finally, we study the order statistics from the Type I generalized logistic distribution in detail. We derive exact and explicit expressions for the means, variances and covariances of order statistics in terms of the gama function and its successive derivatives. Percentage points and modes of order statistics are also discussed

ISSN:0361-0918    

DOI:10.1080/03610918808812648

出版商:Marcel Dekker, Inc.    

年代:1988

数据来源: Taylor

摘要:

In this paper, we first tabulate the means, variances and covariances of order statistics from the Type I generalized logistic distribution for b = 1.0 (0.5) 5.0 (1.0) 8.0 and for sample sizes up to fifteen. These quantities have been used in obtaining the necessary coefficients for the best linear unbiased estimators of the location and scale parameters, and the variances and covariance of these estimators. We also present some examples which illustrate the importance of all these tables in the contexts of estimation and robustness.

ISSN:0361-0918    

DOI:10.1080/03610918808812649

出版商:Marcel Dekker, Inc.    

年代:1988

数据来源: Taylor

摘要:

An extension of the Dunnett test for comparing k experi-mental groups with a single control group is presented for the multivariate case when the control group may contain as many as seven times the cases contained in each experimental group. test, based on coorrelated Hotelling T2statistics, uses a simple adjustment to the equal-n critical values to arrive at values which are slightly conservative over a wide range of unbalanced designs. A real-world application is presented.

ISSN:0361-0918    

DOI:10.1080/03610918808812650

出版商:Marcel Dekker, Inc.    

年代:1988

数据来源: Taylor

摘要:

We study the effects of truncation on the performance of an open matrix–at–a–time sequential sampling procedure [d], described by Bechhofer and Goldsman (1987a) for selecting the normal population which has the largest mean, when the variances of the populations are known and equal. This procedure which was devised for 2–factor experiments with no interaction between the factor–levels is a generalization of the basic procedure [d] proposed by Bechhofer, Kiefer and Sobel (1968) for single–factor experiments. The performance of a truncated version [d] of [d] is compared to that of [d] as well as to [d] the single–stage procedure of Bechhofer (1954). The performance characteristics studied for the sequential procedures include the achieved probability of correct selection (P{CS})

ISSN:0361-0918    

DOI:10.1080/03610918808812651

出版商:Marcel Dekker, Inc.    

年代:1988

数据来源: Taylor

摘要:

In this paper we give an alternative to show the relationship between t and Beta using generalized logistic distribution. An expression for t-cumulative probabilities is then derived by evaluating the incomplete beta integral. An algorithm is also given for computing the t probabilities. This algorithm seems to have some desirable numerical properties.

ISSN:0361-0918    

DOI:10.1080/03610918808812652

出版商:Marcel Dekker, Inc.    

年代:1988

数据来源: Taylor

摘要:

This paper discusses the risk of the linear empirical Bayes p =â+ß y of the binomial parameter p, where y is the observed proportion of successes in the current trial, âand ßare functions of the estimated prior parameters. When â and ß are non linear functions of the observations, two analytical approximate risks of p, assuming a beta - binomial model, have been developed

ISSN:0361-0918    

DOI:10.1080/03610918808812653

出版商:Marcel Dekker, Inc.    

年代:1988

数据来源: Taylor

摘要:

High-to-low-dose extrapolation is a very important part of the quantitative cancer risk assessments involved in the regulation of chemical exposures. Currently, these extrapolations usually involve the multistage model and a likelihood-ratio-based confidence limit procedure; however, both have some critical practical weaknesses. Four alternative confidence limit procedures based on recent bootstrapping research are investigated. Three procedures provide good coverage and lessen the tendency for the confidence limits to be several orders of magnitude from the quantity being bounded. In addition, these procedures can be readily applied to the emerging successors to the multistage model.

ISSN:0361-0918    

DOI:10.1080/03610918808812654

出版商:Marcel Dekker, Inc.    

年代:1988

数据来源: Taylor

摘要:

Let T = x1+ x2+ … + xnwhere x1,…, xnare independent and identically distributed randon variables each having a poission distribution with mean θ. It is proved that each of the estimators ((n+k−1)/(n+k))T, k = 0,1,2,… is admissible for estimating exp(−θ) when the loss function is squared error.

ISSN:0361-0918    

DOI:10.1080/03610918808812655

出版商:Marcel Dekker, Inc.    

年代:1988

数据来源: Taylor

摘要:

Bayesian forecasting methods are applied to the analysis of repeated surveys under a dynamic exponential family superpopulation model. It is assumed that the parameters of the superpopulation model at each time period follow a stochastic model. Some numerical examples are provided to illustrate the behavior of the suggested predictors.

ISSN:0361-0918    

DOI:10.1080/03610918808812656

出版商:Marcel Dekker, Inc.    

年代:1988

数据来源: Taylor