Suchergebnisse

Confronted with entangled behavior, unknown interdependencies, masses of qualitative & quantitative variables, & bad data, many soc sci'ts are turning toward factor analysis to uncover major soc & internat'l patterns. The problem of communicating their results is formidable, however, because many of the scholars are unfamiliar with this type of analysis. The aim is to draw a conceptual map of the method & clarify the technical paraphernalia so that consumers of factor studies can interpret them usefully. Topics include the relationship of factor analysis to sci'fic method; the factor model; interpreting factor tables; & factor rotation. Sample internat'al data are used as illustration, & a Bibliog of factor analysis in conflict & internat'al relations is included. AA.

Confirmatory Factor Analysis (CFA) is used for four major purposes: 1) psychometric evaluation of measures; 2) construct validation; 3) testing method effects; and 4) testing measurement invariance (e.g. across groups or populations). With an easy-to-follow overview of the method, step-by-step guide to creating a CFA model, and clear guideline to requirements for using CFA, this book will be ideally suited for readers who plan to conduct CFA analyses, but want a brief, non-technical introduction to the topic to get them started before getting into the more detailed and technical literature, as well as readers who do not plan to conduct CFA analyses, but want to be knowledgeable consumers of research literature that uses CFA.

Factor analysis came into being around 1900 in the field f psychology to explain theories of human ability. Several methods of factor analysis exist; but according to Harman (1967) principal component factor analysis is unique in the mathematical sense, therefore, quite often the preferred method. The centroid method is computationally easier, and it gives close approximations to the principal component method on some data sets. An example of this is shown in Appendixes B and F by comparison. Factor analysis is being used in many fields. A few of the fields are sociology, meteorology, political science, medicine, geography, business, economics, ecology, soil science, and geology. The following are three specific examples. In meteorology, White (1958) found that factor analysis could reduce considerably the number of variables in his study of sea-level pressure forecasting. In this study, there were 42 original variables. With 5, 10, and 20 underlying variables, White was able to account for 75.51, 90.70, and 97.37 per cent of the original variance, respectively. In ecology, Orloci (1966) found that factor analysis could be used to reduce the number of variables in his study of vegetation on Newborough Warren, Anglesey. In this study, there were 101 original variables. With three underlying variables, Orloci was able to account for 43.98 per cent of the original variance. In the study by White names were not given for the new variables that were found; whereas in the study by Orloci, meaningful names were obtained for the first three factors. Just because a factor accounts for a large portion of the variance does not imply that meaningful names may be readily applied to these factors. In soil science, Lombard (1965) found that factor analysis could be used in his study of citrus irrigations to reduce the number of variables from 12 to 3. These three underlying variables accounted for 99.5 per cent of the variance. Meaningful names were obtained for the three underlying variables. If meaningful names cannot be obtained for the roots or new variables, then factor analysis is not of much value as a statistical tool. Many different programs have been written to perform factor analysis. A few of the existing programs are those by Cooley and Lohnes (1962), Horst (1965), Hurst (n.d., d), in the System/360 Scientific Subroutine Package (1968), and Veldman (1967). This report contains program write-ups and listings for three computer programs, one for principal component factor analysis, one for factor analysis transformation, and another for the centroid method factor analysis. The Principal Component Factor Analysis program will handle up to 50 variables. The Factor Analysis Transformation program will handle up to 50 variables and 15 factors. The Centroid Method Factor Analysis program will handle up to 60 variables. These programs will all run on a 65K byte IBM 360/44 with FORTRAN IV. A card reader, card punch, printer, and one disk or tape are needed.

Introduction -- Creating a CFA model -- Requirements for conducting CFA -- Assessing CFA model fit and model revision -- Use of CFA with multiple groups -- Other issues in CFA -- Glossary

The principal-factor solution is probably the most widely used technique in factor analysis and a relatively straight forward method to determine the minimum number of independent dimensions needed to account for most of the variance in the original set of variables. The principal components approach to parsimony was first proposed by Karl Pearson (1901) who studied the problem for the case of nonstochastic variables, and in a different context. Hotelling provided the full development of the method (1933) and Thomson (1947) was the first to apply it to the principal factor analysis. This method was first developed to deal with problems in psychology but has since been applied in fields as varied as sociology, meteorology, economics, biometry, political science, medicine, geography and business.