摘要:

The estimation of a non‐parametric measure of signal discriminability is discussed and its behaviour investigated under the following conditions: (a) when the ‘trace’ distributions for the two signals being compared differ only by a shift parameter, Λ (b) when the probability density function of the trace is a monotone likelihood ratio; and (c) when the subject bases his response on a likelihood‐ratio rule, with a constant criterion. In this case the measure of discriminability is an increasing function of the shift, A. This monotonicity also holds for the two‐response experiment under rather more general conditions. Data from such an experiment are used as

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1973.tb00501.x

出版商:Blackwell Publishing Ltd    

年代:1973

数据来源: WILEY

摘要:

A formal analysis is made of some conclusions by Anger (1956) and Morse (1966), based on the results of Anger's experimental research into the effects of reinforcement schedules on the distribution of times between successive responses. This analysis shows that the conclusions are fundamentally incorrect.

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1973.tb00502.x

出版商:Blackwell Publishing Ltd    

年代:1973

数据来源: WILEY

摘要:

The fair betting game between two players requires one player to propose the proportional distribution of the total stake over two sides of a bet on the outcome of an uncertain binary event. The other player then chooses his most preferred side of the bet. When the first player has no reason to suppose that his relevant degrees of belief differ from his opponent's, the former's best strategy is to propose such a bet that its two sides have equal subjectively expected utility, or SEU, from which subjective probabilities can be directly inferred, if utilities are known or assumed linear in money. The mathematical properties of this game and their possible effectiveness in bringing about honest probability assessments are examined. Under specified conditions the game amounts to an admissible procedure for probability assessment, characterized by apeakedmaximum SEU for optimal estimation. Two variants, the complete and simplified versions, of the game are discussed and their properties compared. The effects of uncertainty about one's true subjective probability, under a rational as well as under a partially irrational strategy, are briefly investigated. Two possible applications of the method to the assessment of continuous probability distributions are suggested.

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1973.tb00503.x

出版商:Blackwell Publishing Ltd    

年代:1973

数据来源: WILEY

摘要:

Since most of the literature on the extension of Rasch's (1961) model to multiple‐ choice questionnaires is only available in Danish, this paper gives a short introduction to the model. The main body of the paper, however, is concerned with the conditional maximum‐likelihood estimation of the item parameters. The theory is applied to data from a questionnaire used to measure job satisfaction in Danish factories. Finally, the problem of reducing the dimensionality of the parameters is discus

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1973.tb00504.x

出版商:Blackwell Publishing Ltd    

年代:1973

数据来源: WILEY

摘要:

Although the calculated reliability of a test based on a small sample of subjects is an estimate of the population reliability and, hence, is subject to sampling fluctuation, little application has been made of statistical inference techniques to the reliability coefficient. The need for such techniques was recognized as early as the 1940s, and a few attempts have been made to introduce analysis of variance (anova) procedures and to relate the sampling distribution of reliability coefficient estimates to the well‐knownFdistribution (e.g. Jackson&Ferguson, 1941; Ebel, 1951). It was, however, Kristof (1963) who first presented a rather complete sampling theory of reliability estimates and a method to apply it. Feldt (1965) derived similar results based on ananovamodel and applied it to a binary item test case. The above sampling theories were based on normality assumptions regarding the true and error score distributions. It is, however, conceivable that real data will not always meet the rigorous assumption of normality, and yet very little is known about the effects of non‐normality on the sampling distribution of reliability estimates. The present paper contains the results of an investigation of the effects of non‐normality under a mixed modelanova. Both analytical and computer simulation methods are

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1973.tb00505.x

出版商:Blackwell Publishing Ltd    

年代:1973

数据来源: WILEY

摘要:

The Bonferroni inequality can be used to derive a value of Student'st(or a unit normalz) for confidence intervals, and guarantees that the probability is at least 1 — a that all intervals formcontrasts, chosen in advance, cover the true values of their contrasts. An attempt has been made to extend this technique to contrasts chosenpost hoc.It is shown that this extension is not generally vali

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1973.tb00506.x

出版商:Blackwell Publishing Ltd    

年代:1973

数据来源: WILEY

摘要:

When measurements of the same variates are made on individuals sampled from different populations, factor analysis may be used to examine whether there is evidence of common underlying factors. A factor model suitable for this situation is discussed and two methods given for obtaining the estimated loadings of the common factors from the loadings obtained from separate factor analyses. The first method is appropriate when all the factors are common to all the populations; the second when some factors are specific to certain populations or groups of populations. Estimators are given of the factor mean vectors and covariance matrices in the various populations and, with the second method, indications of the populations in which each factor operates. An iterative procedure for obtaining the maximum‐likelihood estimates is explained in detail and some tests of significance are mentioned. Possible rotations of the common factors are discussed and the results of applying these methods to some numerical data are reporte

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1973.tb00507.x

出版商:Blackwell Publishing Ltd    

年代:1973

数据来源: WILEY

摘要:

This paper develops a model for factor analysing scores onntests administered toNpersons onpseparate occasions. It allows factor scores and factor loadings to vary, but assumes the factors to be mutually orthogonal within, as well as between, all occasions. These assumptions are generally sufficient to specify unique rotations on each separate occasion. A worked example is given.

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1973.tb00508.x

出版商:Blackwell Publishing Ltd    

年代:1973

数据来源: WILEY

摘要:

Suppose that we are utilizingkdifferent psychological tests, each having one subtest in common. Of particular concern is the hypothesis that these tests are parallel with respect to their means and/or covariances. A complete hierarchy of hypotheses for this situation has been developed. For example,Hm'veis the hypothesis that the tests are parallel only with respect to the means of the common test, but with respect to the covariances of both tests. (The prime indicates equality for the common test only.) This hypothesis might be tested againstHvc, the hypothesis of parallelism with respect to the covariances. Other hypotheses considered areHmvc,Hm'vc,Hm'vcandHvc. Maximum‐likelihood estimators under the various models (and under the assumption of normally distributed test scores) have been obtained, as well as the related likelihood‐ratio statistics. Approximate distributions of the likelihood‐ratio statistics are worked out, so that the tests can be applied, and an example of their use is pro

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1973.tb00509.x

出版商:Blackwell Publishing Ltd    

年代:1973

数据来源: WILEY

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1973.tb00510.x

出版商:Blackwell Publishing Ltd    

年代:1973

数据来源: WILEY