摘要:

An error by the author in comparing Luce's choice axiom with choice‐reaction data is here corrected. After a brief description of the experiment, a likelihood‐ratio procedure is developed and the data re‐analysed, with attention to why the present analysis corrects the previous error. The axiom appears inconsistent with these data, possibly because the set of stimuli used had a natural physical structure. This result is consistent with other psychophysical studies and is germane to the bounds of applicability of the choice

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1977.tb00734.x

出版商:Blackwell Publishing Ltd    

年代:1977

数据来源: WILEY

摘要:

Based on a given asymmetric proximity function, a two‐stage computational heuristic for sequencing a set of objects along a continuum is presented and illustrated with the type of example common in the paired‐comparison literature. The first stage, defined by the pairwise interchange of objects, is intended to generate reasonably good orderings from randomly chosen initial starts; the second stage can be considered to be a refinement phase and depends on a general condition for an optimal solution suggested by Younger involving the interchange of consecutively placed groups of objects. An extension of the heuristic is also described that allows the selection and sequencing of a subset of all the possible obje

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1977.tb00735.x

出版商:Blackwell Publishing Ltd    

年代:1977

数据来源: WILEY

摘要:

Results given by Sohönemann&Steiger (1976) are related to theory given by McDonald (1975) and McDonald&Burr (1967). From the relation between split‐half estimators of a common factor, it is shown that the factor is uniquely defined in a behaviour domain by a set of marker variables, if it is defined by them at all. Regression components obtained from split‐halves do not, in general, have higher correlations than factor score estimators, hence there is no reason to prefer regression components to common factors on grounds of determi

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1977.tb00736.x

出版商:Blackwell Publishing Ltd    

年代:1977

数据来源: WILEY

摘要:

From a Bayesian framework the classical or likelihood definition of model identification is shown to be too restrictive since there exist some classically unestimable and unidentifiable models that have unique Bayes estimates. Consequently, another definition of identification is developed. Finally, the general issue of model complexity and estimation techniques is discussed.

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1977.tb00737.x

出版商:Blackwell Publishing Ltd    

年代:1977

数据来源: WILEY

摘要:

A method based on Jöreskog's approach to the analysis of covariance structures is applied to the genotype‐environment analysis of the covariation of four aspects of impulsiveness in male and female monozygotic and dizygotic twins. The data are consistent with a simple model which assumes additive gene action, random mating and environmental effects within families. Further, it is shown that genotypic factor loadings can be regarded as constant multiples of corresponding environmental loadings. Trait‐specific sex interactions are detected which suggest that some mechanism of sex‐limitation could contribute to specific var

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1977.tb00738.x

出版商:Blackwell Publishing Ltd    

年代:1977

数据来源: WILEY

摘要:

It is pointed out that both the degrees of freedom in the chi‐square asymptotic likelihood ratio test and the local identifiability of the parameters in the model can be determined unambiguously from the ranks of certain matrice

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1977.tb00739.x

出版商:Blackwell Publishing Ltd    

年代:1977

数据来源: WILEY

摘要:

The attenuation paradox refers to the increase in test validity that accompanies increasing test reliability up to a point beyond which validity decreases with further increases in reliability. It is shown that, for perfectly discriminating items, this phenomenon can occur regardless of the distribution of underlying ability. A numerical example is also given for a class of non‐perfectly discriminating items and a rectangular distribution of abilit

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1977.tb00740.x

出版商:Blackwell Publishing Ltd    

年代:1977

数据来源: WILEY

摘要:

The way in which poorly conditioned matrices can adversely affect the stability of estimates of partial regression coefficients is described. It is noted that in the absence of measurement error in the independent variables relative stability can be achieved by the ‘ridge regression’ method, but when the margin of error is considerable, as is generally the case in social science research, poorly conditioned matrices of an extreme kind occur only infrequently. Whether or not they occur the method of estimating regression coefficients which employs the basic model of factor analysis effectively circumvents inherent difficult

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1977.tb00741.x

出版商:Blackwell Publishing Ltd    

年代:1977

数据来源: WILEY

摘要:

The Kruskal‐Wallis and normal scores non‐parametric tests of location equality are compared to the parametric analysis of varianceFtest. In addition to quantifying and manipulating degrees of variance heterogeneity due to combining heterogeneous variances with unequal sample sizes, the tests are compared for their sensitivity, as well as rates of Type I error, under varying conditions of mean variability when sampling from Gaussian and exponential distributions. Even though the asymptotic literature (Pratt, 1964; Puri, 1964) favours the normal scores test, it is found that the Kruskal‐Wallis test is preferable to the normal scores test while the choice between the Kruskal‐Wallis and analysis of variance tests depends upon populatio

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1977.tb00742.x

出版商:Blackwell Publishing Ltd    

年代:1977

数据来源: WILEY

摘要:

The problems of using the Kruskal‐Wallis (KW) test and two equivalent forms of the normal scores (NS) test in analysing data arising from a continuous distribution are discussed for the cases where the data have been rounded to the nearest integer and first decimal place respectively. When the distribution functionGis normal, the NS test is asymptotically slightly more efficient than the KW test, whereas whenGis double exponential, Cauchy or logistic, the KW test is somewhat more efficient. Even whenGis exponential or rectangular, the NS test is only slightly more efficient than the KW test. This is quite surprising by comparison with the continuous case where the NS test is infinitely more efficient than the KW test. Nor do these conclusions need modification for the cases where the populations differ both in location and in scale and where the differences among the location parameters and among the scale parameters tend to zero as the sample sizes tend to infinity, providedGis normal, double exponential, Cauchy or logistic. When homogeneity of variances obtains, the results follow from those of Conover (1973), but with heterogeneous variances his techniques have to be modified as in Goldsmith&Padmanabhan (1976). The limitations of these results and problems for further investigation are explaine

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1977.tb00743.x

出版商:Blackwell Publishing Ltd    

年代:1977

数据来源: WILEY