HYBALL FACTOR-ANALYSIS

Download: FACTOR-ANALYSIS (Apr 8,2005)


26 March 2002

Hyball is an integrated collection of stand-alone DOS-operated interactive routines for exploratory factor analysis (EFA) having many advanced features afforded by few if any commercial data-analysis packages. Its core is a program for rotating to oblique simple structure a matrix of factor coefficients previously extracted from the correlations within an array of data variables. Its innovative solution logic is especially proficient at detection of hyperplanes, and facilitates discovery of what may or may not prove to be a diversity of interpretively provocative pattern alternatives latent in the factored data. Provisionally favored alternative solutions are collected in archives pending comparison study of these by means of additional Hyball resources designed to make differences among alternative factor positionings finely perspicuous. These resources include comparison of solutions varied in their number of extraction factors and/or factoring of covariances within just a subset of the variables in the raw-data collection at hand, allowing study of how robustly certain factors can be recovered under variation in these solution conditions.

Other programs in the Hyball package variously compute item correlations from raw data with optional imputation of missing scores, extract initial factor patterns from these by choice from an assortment of established factoring methods, impose a causal-path structure on blocks of extracted factors -- an EFA counterpart of modern Structural Modeling (SEM) approaches to multivariate data analysis -- if there is prior reason to posit that the data variables are indicators of latent variables so ordered, and create practical item scales for factor estimation. Provision is also included for bootstrap estimation of sampling noise in the terminally rotated factor patterns.


Save this ZIP-file to whatever subdirectory you create to be Hyball's home (an easy name for that is "HYBL" if you have no other preference) and either clicking on it there to open (run) while still in Windows or, after going to that location in a DOS window, running HYPAK as a DOS command. That will deliver three more ZIP ZIP-files and an ASCII textfile, READ_1ST.TXT, that advises you on what to do next. From there, all your work with Hyball needs to be done within a DOS window that should be expanded to full screen.


REFERENCES
  • Rozeboom, W. W. (1991). HYBALL: A method for subspace-constrained oblique factor rotation. Multivariate Behavioral Research, 26, 163-177.
  • Rozeboom, W. W. (1992). The glory of suboptimal factor rotation: Why local minima in analytic optimization of simple structure are more blessing than curse. Multivariate Behavioral Research, 27,585-599.