PROXIMITY INDICATORS IN AUTHOR CO-CITATION ANALYSIS: a comparative study between the Pearson's correlation coefficient and Salton's cosine

Abstract

This study aims to present and analyze Pearson Correlation Coefficient (r) and Salton's Cosine (CS) as proximity indicators to Author Co-citation Analysis and compare the contribution of these two indicators to visualize proximity among co-cited authors, evaluating the correlation between the indicators and identifying clusters of co-cited researchers formed from the two indices, using published articles on the topic Metric Studies from BRAPCI database. From a previous research, we retrieved the matrices with absolute frequencies of co-citation of the 38 most cited authors, normalized by Salton's cosine. Then the matrix was yielded with the values of the Pearson's correlation among the co-cited authors, using the SPSS software. The indices were analyzed, especially Salton's cosine and the Pearson's correlation. The correlation among the 703 pairs from both indices were calculated, represented by a scatterplot of values. From each matrix-Salton's Cosine and Pearson's Correlation-, two analyzes of clusters were performed, generating two dendograms, which were used to the comparative analysis of the clusters. In conclusion, we highlight that coauthors have high indices of CS and r. Moreover, we observed a trend toward low values for CS paired with low values of r. We concluded, therefore, that although CS measures local proximity and r measures global proximity, these indices show the same trends in relation to proximity.