|
Essentials of Feature Pattern Analysis
Hubert Feger, Karl Christoph Klauer
Summary
Feature Pattern Analysis (FPA) is a model to explain vectors of
dichotomous or polytomous variables by the lowest contingency fitting
the data. FPA represents the data equivalently either as regions of a
geometric space of lowest dimensionality or as a consistent set of
logical prediction rules with a minimum of predictors. It is shown
that Guttman scaling and parallelogram analysis are special cases of
FPA, and both are generalized to the multidimensional case. This paper
specifies procedures and criteria for FPA, scalogram and parallelogram
analysis in the one and multidimensional case, including an
approximate procedure for data with error. It provides examples for
diagnostic classification, attitude measurement, and the analysis of
data from the Semantic Differential.
Key words: Co-occurrence data, Guttman scaling,
parallelogram analysis, Semantic Differential, logical prediction
rules, discriminant analysis, quantification
Dr. Hubert Feger
Department of Psychology
Free University of Berlin
Habelschwerdter Allee 45
D-14195 Berlin
Germany
    
|
PABST SCIENCE PUBLISHERS
Lengerich, Berlin, Riga, Rom, Wien, Zagreb
|
|
|
