Citation analysis: procedures used

In the next phase, the co-dtation matrix needed to be constructed and visualized. For the construction of the matrix, references were either extracted automatically (ISl) or manually (EBSCO) for each publication included in the analysis. As the entries in die reference lists included different citation styles or misspellings, all references needed to be normalized. I then excluded those references which were not part of the dataset, i.e. those which were not included in the 111 papers. By this procedure many references were excluded from the analysis, but the matrix could be constructed more effidendy. The focus of the citation analysis was rather on the connection of papers within the field of user innovation than on the epistemological foundations thereof (cf. Raasch et al. forthcoming). Thus it was sufficient to include only papers that focus on the topic of user innovation, and not its intellectual ancestry. In the next step I constructed the citation matrix. This 111x111 matrix included all articles and indicated which other publications from the pool were cited. Publications that were not died at all and did not co-dte others were deleted from the matrix, as they were obsolete for the analysis. This yielded a list of 100 publications. In the last step the co-citation matrix was built I exduded papers that were not co-cited (46 publications) from the analysis. The doseness measure used in this study (CoCit score) was calculated. The CoCit score of two documents (A and B) ranges from 0 to 1 and can be calculated as follows (Gmiir 2003) ... 

Here we adopt the more limited objective of using a rudimentary form of citation analysis to assess the shifting distribution of attention to American chemical research by German and British chemists. (Sampling procedures are discussed in Appendix C.)... 

---