摘要:

The common problem of comparison of counts arising from Poisson distributions is discussed and exemplified for any number of Poisson means and any values of associated opportunties. Confidence interval estimates when there are two counts are exemplified. The relationship between the nature of the alternative hypothesis and the power of the test is illustrated.

ISSN:0022-4065    

DOI:10.1080/00224065.1987.11979062

出版商:Taylor&Francis    

年代:1987

数据来源: Taylor

摘要:

The use of boxplots in place of single points in a quality control chart can provide an effective display of the information usually given in X̄andRcharts, show the degree of compliance with specifications and identify outliers. An example from a process evaluation illustrates the use of the proposed boxplot chart and compares it to the customary X̄andRcharts. To achieve a degree of robustness, the present paper compares several resistant measures of location and dispersion in terms of their efficiency for data from the normal, double exponential and“one-out”distributional situations. The results indicate that the mean and range of the middle half of the sample (suitably defined) combine to yield a boxplot chart with good statistical properties. For each pair of location and dispersion measures studied, a table gives the necessary factors to construct control limits.

ISSN:0022-4065    

DOI:10.1080/00224065.1987.11979063

出版商:Taylor&Francis    

年代:1987

数据来源: Taylor

摘要:

Existing sampling plans are not very efficient in situations where the fraction of nonconforming items is near 0.0001. In such cases the statistician must resort to 100% inspection, which still will not guarantee 100% conforming product since inspection error can introduce an error almost as great as the original fraction nonconforming. Skip-lot sampling is rarely an acceptable alternative to customers, who usually prefer inspection of every shipped lot. This paper suggests an inspection policy combining the ideas of several sampling strategies and guaranteeing against both undesirable levels of Average Outgoing Quality (AOQ) and poor quality on a lot-to-lot basis. The method is based on observing the number of conforming items between nonconforming ones and shifting to 100% inspection when certain criteria are not met.

ISSN:0022-4065    

DOI:10.1080/00224065.1987.11979064

出版商:Taylor&Francis    

年代:1987

数据来源: Taylor

摘要:

The laboratory ranking test is a nonparametric procedure that W. J. Youden recommended in 1963 for identifying outlying laboratories in collaborative studies. Though potential problems with the performance of this test have been noted by Youden and others, its use still is being promoted. One problem with the test is that its critical values were developed for a different application and do not allow for the between-laboratory variation ordinarily observed in collaborative studies. This paper presents a quantitative evaluation of the impact of between-laboratory variation and other factors on the performance of the laboratory ranking test. The results document some of the test's shortcomings, including its propensity to identify unexceptional laboratories as outliers.

ISSN:0022-4065    

DOI:10.1080/00224065.1987.11979065

出版商:Taylor&Francis    

年代:1987

数据来源: Taylor

摘要:

ANSI/ASQC Z1.4 is a widely used attributes acceptance sampling system. The latest revision of the standard makes optional the use of limit numbers for reduced inspection. This paper presents a Markov model of the sampling system without limit numbers. The model is valid for all single, double and multiple sampling schemes. Closed form, analytical results are developed for various performance measures. The maximum average run length on normal inspection is found to be only 12.7 lots, which leads to excessive switching between plans when quality is worse than the AQL.

ISSN:0022-4065    

DOI:10.1080/00224065.1987.11979066

出版商:Taylor&Francis    

年代:1987

数据来源: Taylor

摘要:

To design a random effects one-way analysis of variance experiment one needs to determine both a, the number of treatments that are to be included in the experiment, as well asn, the number of times that each treatment is to be observed. In the fixed effects experiment only the number of times that each treatment is to be replicated needs to be determined since the number of treatments of interest is already fixed. We consider the problem of designing a random effects experiment where the total number of observations, N, is fixed and the power of theF-test is to be maximized. This problem is simple when one knows apriorithe value of the error standard deviation. We present a method which allows the experimenter to state a distribution for the error standard deviation. Our objective is then to find the experimental design which maximizes the expected power.

ISSN:0022-4065    

DOI:10.1080/00224065.1987.11979067

出版商:Taylor&Francis    

年代:1987

数据来源: Taylor

摘要:

This paper presents a FORTRAN computer program for allocating observations in random effects balanced one-way ANOVA experiments when the total number of observations is fixed.

ISSN:0022-4065    

DOI:10.1080/00224065.1987.11979068

出版商:Taylor&Francis    

年代:1987

数据来源: Taylor

ISSN:0022-4065    

DOI:10.1080/00224065.1987.11979069

出版商:Taylor&Francis    

年代:1987

数据来源: Taylor

ISSN:0022-4065    

DOI:10.1080/00224065.1987.11979070

出版商:Taylor&Francis    

年代:1987

数据来源: Taylor

ISSN:0022-4065    

DOI:10.1080/00224065.1987.11979071

出版商:Taylor&Francis    

年代:1987

数据来源: Taylor