Quasi-correspondence analysis on scientometric transaction matrices

In principle, a scientometric transaction matrix can be modelled by assuming that the number of transactions is the result of independent row and column contributions. More often one is primarily interested in the cross-structural relations between the participating entities, whereas the row and column margināls are of lesser or no importance. The values of the residuals after fitting an independence model to a complete transaction matrix can be analyzed by correspondence analysis to investigate the structure of the transactions between the rows and columns, after correcting for their marginal frequencies. Recently a modification of correspondence analysis has been developed, quasi-correspondence analysis, which seems quite suitable for the analysis of citation-based transaction matrices which are incomplete or in which the incorporation of certain transactions may seem inappropriate. An illustration of both data analysis-techniques will be given using a journal-to-journal citation matrix.