摘要:

This paper reviews the developments in generalizability theory from 1973 to 1980. The first section presents a sketch of generalizability theory. The second section reviews theoretical contributions about (1) problems associated with estimating variance components, including sampling variability and negative estimates, (2) fixed facets, (3) criterion‐referenced measurement, (4) symmetry, (5) multivariate generalizability, and (6) sampling in observational measurement. The final section presents an illustrative application of generalizability theory, including univariate and multivariate generalizability analyses of balanced and unbalanced designs, and Bayesian estimation of variance component

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1981.tb00625.x

出版商:Blackwell Publishing Ltd    

年代:1981

数据来源: WILEY

摘要:

Previous work on representing proximity data by tree structures is reviewed. The authors then present two paradigms for data collection and develop appropriate measurement structures for non‐numerical data. Two theorems on representing these data by rooted or unrooted trees are given and their relationship to the more conventional representation by a dissimilarity index is discusse

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1981.tb00626.x

出版商:Blackwell Publishing Ltd    

年代:1981

数据来源: WILEY

摘要:

A Bayes decision approach is adopted to give predictive models for the behaviour of an inspector or inspectors searching for an unlikely fault. The associated probability of the detection of such an unlikely fault is found and is related to the inspector's prior beliefs, losses and personal model when he or she is in various environments. These theoretical probabilities of detection that assume ‘sensible’ action from the inspector are found, generally, to be dramatically smaller than the probabilities obtained from the assumption of efficient search. Thus the paper explains why events with associated minute ‘probabilities’ seem to occur regularly in p

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1981.tb00627.x

出版商:Blackwell Publishing Ltd    

年代:1981

数据来源: WILEY

摘要:

The aim of this paper is twofold: Firstly, a mathematical procedure is described for estimating interscale relations between category and magnitude scales. The computer program based on this procedure is designed for replacingad hoctechniques that are widely employed in exploring between‐scale relations. Secondly, the new method is applied to empirical data from a large‐scale survey, and a range‐dependent model for the category‐magnitude relation is p

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1981.tb00628.x

出版商:Blackwell Publishing Ltd    

年代:1981

数据来源: WILEY

摘要:

Goldstein (1979) showed that interdependence of Bernoulli data can lead to either overestimation or underestimation of odds depending on the type of interdependence, and inferred that conservative behaviour is not a necessary outcome of interdependence. It is argued here that the effect of data interdependence on posterior odds in the real world cannot be determined by pure statistical reasoning. Logical arguments suggest that interdependence which requires conservative revisions is most typical. Thus, a Bayesian processor acting on the assumption that data are conditionally independent is likely to end up with inflated posterior odds.

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1981.tb00629.x

出版商:Blackwell Publishing Ltd    

年代:1981

数据来源: WILEY

摘要:

In large public examinations many examiners are required to mark the scripts. These examiners are to be calibrated by requiring them all to mark a sample of common scripts. Paul (1979) used an additive model and a linear relationship model to describe the examiners' behaviour in marking the scripts, and gave estimates of the parameters from a sampling theoretic approach. In this paper Bayesian estimates of the parameters of the models are given. Real examination data and simulated data are analysed and the conclusion of Paul (1979) that there is very little to choose between the models is confirmed. It seems that the additive model is adequate.

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1981.tb00630.x

出版商:Blackwell Publishing Ltd    

年代:1981

数据来源: WILEY

摘要:

Wilcox (1979) has presented in this journal a model for achievement testing where a population of examinees is described in terms of a domain of skills. In most of his paper each examinee knows a proportion ζ of all items from this domain, and has a probability β of guessing the correct response given that he/she does not know it. Across examinees, the parameters ζ and β (1 — ζ) have a bivariate Dirichlet distribution. Per examinee, ζ and β are estimated from the results on pairwise equivalent item pairs. The present note explores the feasibility of the assumptions and the usefulness of the results of Wilcox's paper. It thus attempts to show the obstacles between the present psychometric model and a fully satisfactory treatment of

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1981.tb00631.x

出版商:Blackwell Publishing Ltd    

年代:1981

数据来源: WILEY

摘要:

Commenting on a paper I published in this journal (Wilcox, 1979a), Molenaar (1981) has raised some questions about the usefulness and feasibility of measuring achievement with latent structure models. In the last two or three years, considerable progress has been made regarding the issues mentioned by Molenaar. The purpose of this note is to indicate the progress that has been made, to describe alternative solutions that have been recently proposed, and to comment on some of Molenaar's suggestions on how the model might be improved.

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1981.tb00632.x

出版商:Blackwell Publishing Ltd    

年代:1981

数据来源: WILEY

摘要:

Fhanér (1974) proposed an approach to measuring achievement where the binomial error model is assumed, and where the goal is to determine whether an examinee's per cent correct true score is above or below a known constant. Wilcox (19806), as well as van den Brink&Koele (1980), point out that a substantially larger number of items might be required when guessing is incorporated into Fhanér's solution. The purpose of this brief note is to derive the exact sampling distribution of a closed sequential procedure that solves the problem considered by Fhanér. We then show that the probability of a correct decision under the new procedure is exactly the same as it is when Fhanér's procedure is applied. In addition, the number of observations under the closed sequential procedure is always less than or equal to the number required under the fixed sample size approach. In some cases, the number of observations is considerably l

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1981.tb00633.x

出版商:Blackwell Publishing Ltd    

年代:1981

数据来源: WILEY

摘要:

In experimental psychology it is usually difficult to show that populations sampled meet the requirements for the use oftorFtests, or even that they are similar to populations sampled in Monte Carlo experiments designed to demonstrate the robustness of these parametric tests. Consequently, a test which makes weaker requirements without sacrificing power or versatility should be preferred. It is shown that this is true of modified approximate randomization tests, which, like the randomization tests on which they are based, use observed scores to set up a sampling distribution. The tests are versatile, since they can be used on factorial designs of any complexity and the results of Monte Carlo experiments indicate that their power approximates that of theFtest when the assumptions underlying the latter are met and is significantly greater when the populations sampled are made up of two normal distributions with different means. It is concluded that, where adequate computing facilities are available, the approximate randomization test is preferable to anFtest for analysis of variance designs.

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1981.tb00634.x

出版商:Blackwell Publishing Ltd    

年代:1981

数据来源: WILEY