Reliable Component Analysis Through Maximizing Composite Reliability

Reliable component analysis (RCA) provides (a) the weights that construct the most reliable linear composite of a battery's variables, (b) additional sets of weights that give the most reliable composites, uncorrelated with the first and each other, and (c) the reliabilities of each composite. The composites can be used to give factor loadings for the variables on the corresponding components. RCA is a promising method for exploratory factor analysis, because the components' reliabilities can provide a rational basis for the number of factors to rotate. The similarities and differences of component and common factor models are sketched. Artificial and real data are analyzed to illustrate RCA and suggest how its results may differ from those of other methods. Appropriate use of RCA rests on having good reliability estimates. Computations of RCA in a macro computer language are described.