摘要:

Tests based on ranks and the F-test are compared for block designs with n observations per block-treatment combination. Com-parisons are made on level of significance and on power. Rank tests examined include the Friedman as well as those using aligned ranks, weighted ranks, and the rank transformation. It is seen that the performance of these tests in relationship to each other depends on sample size, distribution of the random error term, and the severity of the block effects.

ISSN:0361-0918    

DOI:10.1080/03610918708812607

出版商:Marcel Dekker, Inc.    

年代:1987

数据来源: Taylor

摘要:

A test is proposed for testing the null hypothesis that the median of a distribution is a specified value n0without any assumption about the symmetry of the distribution. The signed-rank test is generally more powerful than the sign test for testing the null hypothesis that the distribution is symmetric around n0, but for our problem it has the disadvantage that it may reject a correct null hypothesis due to the skewness of the distribution. To guard against this possibility, the following procedure is proposed: If the signed-rank test has a highly significant value, reject H0if it has an intermediate value, calculate the value of the sign test statistic and reject H0if this value is significant. It is shown that for suitable criti-cal values this test is robust and has good power properties compared to the sign test.

ISSN:0361-0918    

DOI:10.1080/03610918708812608

出版商:Marcel Dekker, Inc.    

年代:1987

数据来源: Taylor

摘要:

For comparing two competing testing procedures whose power functions cannot be calculated exactly, the Pitman asymptotic relative efficiency is commonly used. This measure offers the advantages of simplicity in computation and interpretation, but is asymptotic in nature. In particular, it cannot distinguish between certain pairs of tests, namely, those with an asymptotic relative efficiency of 1. In this paper, we offer an alternative basis for making finite sample comparisons between tests judged asymptotically equally efficient. The criteria for comparison is based on the ratio of efficacies. Using this new criterion, we consider a series of examples in which two tests with an ARE of 1 are compared for finite n. In each case, either exact or simulated power curves are used to evaluate the effectiveness of this criterion in properly discriminating between the competing test procedures. Lastly, conditions are specified where the proposed criterion should be most effective.

ISSN:0361-0918    

DOI:10.1080/03610918708812609

出版商:Marcel Dekker, Inc.    

年代:1987

数据来源: Taylor

摘要:

This paper deals with six test statistics, of independence of row and column classifications in an(rxc) contingency table. Since their distributions are not easy to obtain, a simulation procedure is used to simulate their percentiles. Four new modified test statistics have been suggested. A power comparison has been conducted and it shows that some of these tests are as good as the usual chi-square, and sometimes they are more powerful than the chi-square test.

ISSN:0361-0918    

DOI:10.1080/03610918708812610

出版商:Marcel Dekker, Inc.    

年代:1987

数据来源: Taylor

摘要:

Because of their nice properties relative to the frequency of occurrence of n-tuples, Tausworthe sequences have been used frequently in recent years as the basis for random number generators. This paper describes a similar result relative to the frequency of occurrence of maximal-length strings of like bits. In particular, for any full-period Tausworthe sequence, the distribution of string lengths is precisely equal to the expected distribution of string lengths for a truly random bit sequence of comparable size.

ISSN:0361-0918    

DOI:10.1080/03610918708812611

出版商:Marcel Dekker, Inc.    

年代:1987

数据来源: Taylor

摘要:

This paper provides five percent significance bounds on critical values of the twelfth or-der analogue of the Durbin-Watsontest. These tables are useful for testing for twelfth order autocorrelation in regression models with monthly data which include either an intercept or a full set of monthly seasonal dummies.

ISSN:0361-0918    

DOI:10.1080/03610918708812612

出版商:Marcel Dekker, Inc.    

年代:1987

数据来源: Taylor

摘要:

This paper develops an iterative estimation procedure for estimating both the parameters and covariances of the seemingly unrelated regression model with contemporaneous covariances. The method is based on the use of an efficient recursive estimation algorithm for the SUR model with known contemporaneous covariances. The components of each observation vector are used one at a time to simultaneously estimate the parameters using a Bayesian “state vector” approach. This is computationally efficient, since the only matrix inversion required is one by one. The covariances are estimated using the residuals from the fitted model. The procedure iterates between estimating the parameters and estimating the covariances until convergence. The algorithm yields the parameter estimator and an estimator for its distribution, thus providing a sound basis for inferential procedures on the parameters.

ISSN:0361-0918    

DOI:10.1080/03610918708812613

出版商:Marcel Dekker, Inc.    

年代:1987

数据来源: Taylor

摘要:

We describe an extension of the recursive algorithm of Howard (1972) and Krailo and Pike (1984) for computation of conditional maximum likelihood estimates in multiple logistic regression models for stratum-matched retrospective studies to the multivariate case involving polytomous disease states. APL procedures, suitable for a desktop computing environment, are provided and analyzed.

ISSN:0361-0918    

DOI:10.1080/03610918708812614

出版商:Marcel Dekker, Inc.    

年代:1987

数据来源: Taylor

摘要:

The problem of testing for exponential regression against a gamma alternative is considered, when it is assumed that the link function for the means is correctly specified. Tests based on the deviance, likelihood ratio, Wald and related statistics are exami-ned, together with some modified tests which incorporate bias cor-rection. Simple chi-square approximations to the powers of the tests are proposed and evaluated by simulation for the case of a logarithmic link.

ISSN:0361-0918    

DOI:10.1080/03610918708812615

出版商:Marcel Dekker, Inc.    

年代:1987

数据来源: Taylor

摘要:

Several kinds of sampling methods (simple random sampling, regular grids, preferential grids, stratified random sampling, and importance sampling) and estimators (polygons of influence and kriging) were used with simulated data to estimate the average level of a pollutant that is highly concentrated around a source. Very precise estimates were given by stratified sampling with a regular grid in each stratum, the grid spacing being reduced in the stratum with the highest concentration. Stratified random sampling and importance sampling gave more variable estimates, but allowed the sampling variance to be estimated from a single sample. Estimates from polygons of influence and kriging were biased, and the kriging variances often differed greatly from the observed sampling variance of the kriging estimates.

ISSN:0361-0918    

DOI:10.1080/03610918708812616

出版商:Marcel Dekker, Inc.    

年代:1987

数据来源: Taylor