摘要:

It is shown that PFA, CFA and AFA are particular cases of a scale‐invariant factoring procedure based on variance ratios of certain weighted combinations of variables. Standard derivations in the literature are shown, in contrast, to have unsatisfactory features. It is suggested that the choice between PFA, CFA and AFA involves relatively independent choices of features of each, and that in most cases CFA is to be preferre

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1970.tb00432.x

出版商:Blackwell Publishing Ltd    

年代:1970

数据来源: WILEY

摘要:

The need to compare factor score estimates across groups differing in age and culture, for example, can be met whenever some variables, factored in all the groups to be compared, are identical.Within the ordinary factor analytic procedures, anequipotentmethod is described, which throws scores from different groups into a combined standard score distribution, and operates on the scores of each group with its own factor estimation weights, but with the matrices for all the groups brought to equipotency.A more satisfactoryisopodicprocedure is also proposed, using covariance factoring, after first reducing artificial metric differences among variables by ‘universe standardizing’ conversions. But a more radical attainment of the isopodic goat is reached if the mean standardized covariance matrix is applied to the raw score deviations of members of all groups, brought into a combined distribution on each varia

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1970.tb00433.x

出版商:Blackwell Publishing Ltd    

年代:1970

数据来源: WILEY

摘要:

The difficulty of finding the second derivatives of the likelihood function has meant that until now it has not been possible to use the usual Newton–Raphson approach to maximum‐likelihood estimation for the factor model. This paper gives a method of finding these second derivatives and shows how they lead to a simple and rapidly convergent numerical method for the estimation of factor model parameters. Some comparisons are made with another recent method due to Jöreskog&Lawley. A useful feature of the method is an asymptotically correct expression for the sampling variances of the estimated parame

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1970.tb00434.x

出版商:Blackwell Publishing Ltd    

年代:1970

数据来源: WILEY

摘要:

An axiom set for similarities is presented together with a set‐theoretic definition of multidimensional similarity consistent with the axioms. A special case of relative similarity judgement, involving the simultaneous comparison of three stimulus pairs, called ‘quadratic similarities’, is empirically illustrated and related to the theory. The goodness of fit of the axiomatic theory is checked for both unweighted and differentially weighted versions. Agreement generally varies between subjects, but in some cases is very good, and is almost invariably improved slightly by weig

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1970.tb00435.x

出版商:Blackwell Publishing Ltd    

年代:1970

数据来源: WILEY

摘要:

A sufficient condition is given that scale values obtained with Gulliksen's (1956) incomplete data solution be unique up to an additive constant.

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1970.tb00436.x

出版商:Blackwell Publishing Ltd    

年代:1970

数据来源: WILEY

摘要:

Empirical investigations suggest that dependency between successive responses in psychophysical experiments may not be uncommon. This paper puts forward a simple Markov model of response dependency. The constant method is briefly reviewed in the light of this model, and is shown to result in unbiased estimates of the psychophysical function. The method of limits is examined in more detail. Estimates of psychophysical function parameters obtained by this method, whether by averaging stopping‐point values or by conducting a probit analysis on the derived data, are shown to be a function of the inter‐stimulus interval chosen, the stopping criterion selected, and the degree of response dependency. However, bias is very small when stopping‐point averages are employed, provided that the inter‐stimulus interval is fairly large compared with the slope of the psychophysical function, and that use is made of a simple correction

ISSN:0007-1102    

DOI:10.1111/j.2044-8317.1970.tb00437.x

出版商:Blackwell Publishing Ltd    

年代:1970

数据来源: WILEY