ERIC Number: ED161891
Record Type: RIE
Publication Date: 1974-Mar
Reference Count: 0
An Examination of Sampling Characteristics of Some Analytic Factor Transformation Techniques.
Skakun, Ernest N.; Hakstian, A. Ralph
Two population raw data matrices were constructed by computer simulation techniques. Each consisted of 10,000 subjects and 12 variables, and each was constructed according to an underlying factorial model consisting of four major common factors, eight minor common factors, and 12 unique factors. The computer simulation techniques were employed to generate one population of factorially simple, and one of complex data, with first the oblique and orthogonal factor models and then the solutions based on several special cases of the general orthomax criterion compared on these characteristics. Results indicated that a significantly better fit to the oblique population factor pattern matrix was realized by the oblique, as opposed to orthogonal, least squares transformations. When the magnitude of the differences between population values was compared with the magnitude of the mean differences for the sample values, the minute population differences had no effect on mean differences for the factorially simple data or the complex data. The authors outline the practical implications of these findings and offer recommendations for future research. (Author/JAC)
Publication Type: Reports - Research
Education Level: N/A
Authoring Institution: Alberta Univ., Edmonton. Div. of Educational Research.