PubMed Health⌕ Search
Home/Archive/PubMed

PubMed · 618970

Principles for identifying structural differences. Some methodological issues related to comparative factor analysis.

W R Cunningham
View original source ↗Export metadata ↓

Abstract

Several recent articles published in the Journal of Gerontology have been concerned with changes in factor structure of various psychological, social, and medical fitness measures. The present article reviews four principles for comparing structures and evaluates several recent articles in terms of these principles. The four principles are concerned with resolving: (1) problems of metric, (2) problems of rotation to similarity, (3) the number of factors problem, and (4) problems related to different extraction or rotational procedures across comparison groups. Utilization of a different metric in each correlation matrix implies that results may be a simple artifact of a different metric for each group. Failure to rotate to similarity and insufficient care with regard to the number of factors problem may result in artificially dissimilar structures. Finally, in comparing results across studies, differences in extraction or rotational procedures across different studies may result in differing results as artifacts of the methods employed.

Explore related subjects

Factor Analysis, StatisticalHumansMetric SystemPhysical FitnessResearchRotationComparative StudyJournal Article

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

W R Cunningham. 1978. Principles for identifying structural differences. Some methodological issues related to comparative factor analysis.. https://doi.org/10.1093/geronj%2F33.1.82

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related citations

PubMed1992

The spectral analysis of hypnotic performance with respect to "absorption".

In factor analyses of the hypnosis scales, the essential result is that the items form a continuous, 2-dimensional fan-shaped pattern. This continuum is referred to as the "spectrum of hypnotic performance." "Special analysis" is introduced as an exploratory procedure which makes use of this notion of continuum or spectrum. Spectral analysis consists of a graphical display of the level of latent correlation between a variable and individual hypnotic performances when the latter are arranged according to their position in the spectrum. The spectral analysis of hypnotic performance with respect to absorption is illustrated using data from a sample of 160 Ss. The results indicate that absorption is more strongly related to difficult hypnotic performances than to easy ones. In particular, illustrative item characteristic curves are presented to show that although easy hypnotic performances do not require the processes tapped by individual differences in absorption, a certain level of absorption is necessary to pass difficult hypnotic items. In addition, a high level of absorption may be sufficient in and of itself for difficult hypnotic performances. These results are discussed in light of some speculations by Shor, M. T. Orne, and O'Connell (1962) and Tellegen (1978/1979) concerning the differential contribution of ability components to performance on difficult hypnotic suggestions. The results are also related to a variety of work in social psychological models of hypnotic performance.

Factor Analysis, Statistical↗
PubMed1992

Implicit personality theory and the five-factor model.

The Big Five have not only been identified in ratings of knowledgeable informants, but also in ratings of strangers, in co-occurrence likelihood ratings of traits, in semantic similarity ratings of trait pairs, and in prototypicality ratings of acts for traits. This article describes the shared and distinctive characteristics of correlations among trait ratings and implicit personality theory and reviews studies that compare the structure of memory-based ratings and of on-line behavior counts. Three hypotheses suggested in the literature to account for these correlations are delineated and discussed: an accurate reflection hypothesis, a distortion hypothesis, and an overlap hypothesis. It is concluded that the distortion hypothesis, has been discredited and that an overlap model best accounts for the available evidence. This implies that traits are real and accurately perceived, provided that the judges have the necessary information.

Factor Analysis, Statistical↗