McCain, K. W. (1990). Mapping authors in intellectual space: a technical overview. Journal of the American Society for Information Science, 41(6), 433-443.
vis_paper
本論文說明作者共被引分析(author cocitation analysis, ACA)的進行步驟與相關技術，ACA的分析流程包括1)選取即將分析的作者集合、2)取得作者的共被引次數、3)建立原始的作者共被引矩陣、4)利用原始共被引矩陣計算相關係數，每一對作者之間以他們與其他作者共被引次數分布的相似程度作為他們之間的接近值、5)對接近值矩陣進行叢集分析(cluster analysis)、多維尺度分析(multi-dimensional scaling, MDS)和因素分析(factor analysis)，產生視覺化圖形、6)解釋與驗證。
Within a given map, the proximity of points representing authors reflects their perceived similarity on some dimension. By examining the distribution of authors and author clusters within the two- or three-dimensional “intellectual space” of a mapped display, other aspects of structure can be described. Clusters of points can be identified with subject areas, research specialties, schools of thought, shared intellectual styles, or temporal or geographic ties. In a factor analysis, factor loadings may demonstrate the breadth or concentration of various authors’ scholarly contributions.
A common sequence of steps in author cocitation analysis is as follows,
1) Selection of the author set: One relatively objective way to identify potentially well-cited authors is to choose those who have many page references in a text, monograph, or collection of review articles.
2) Retrieval of cocited author counts
3) Compilation of raw cocitation matrix
4) Conversion of the raw data matrix to a matrix of proximity values: The creation of a correlation matrix has at least two major advantages. First, for any given pair of authors, the correlation coefficient functions as a measure, not just of how often that pair of authors were cocited (the raw frequency count), but of how similar their “cocitation profiles” are. ... The correlation coefficient also removes differences in “scale” between authors who are highly cited and those who have similar profiles but are less frequently cited overall (Kerlinger, 1973). ... The correlations are defined as measures of similarity: the higher the positive correlation, the more similar two authors are in the perceptions of citers.
5) Approaches to multivariate analysis have been used to display the inter-author relationships in the similarities matrix:
a) In ACA, cluster analysis is used to group authors so as to provide insights into the intellectual organization of a given field. ... The two most popular approaches to cluster formation are called “hierarchical agglomerative” vs. “iterative partitioning” ... ACA research has tended to use the agglomerative clustering approach. The hierarchical agglomerative methods can use the correlation matrix as similarity measures among the authors. Authors are paired, an author is joined to an existing cluster, or two clusters are fused based on their similarity.
b) Multidimensional scaling (MDS) requires as input the same matrix of similarities or dissimilarities among objects as cluster analysis, and the two are often used together. MDS is a set of techniques used to create visual displays- maps -from proximity matrices, so that the underlying structure within a set of objects can be studied. In ACA, the major uses of multidimensional scaling are two-fold -to provide an information-rich display of the cocitation linkages and to identify the salient dimensions underlying their placement. ... Authors heavily cocited (because of their common subject or methodological interests) appear grouped in space. Authors with many links to others tend to be in central positions, while authors weakly linked, or with a few focused ties, will be placed in the periphery. In this way, “central” and “peripheral” research specializations, schools of thought, or other intellectual groupings can easily be seen, ... Dimensions are interpreted based on examination of the author and cluster placements. ... The stress value reported for each solution (usually Kruskal’s Stress I or Stress II) and the proportion of variance explained (R Square in ALSCAL) are indicators of the overall “goodness of fit” of that point configuration.
c) Factor analytic techniques may be used to complement MDS and clustering displays. ... Essentially, they attempt to “explain” the interrelationships observed among the original variables through the creation of a much smaller number of “derived” variables or factors. In ACA, a factor is interpreted by the subset of authors loading on it - i.e., making substantial contributions to its construction. Essentially it reveals their underlying subject matter, as perceived by citers. ... ACA most commonly uses a principal components analysis, with an orthogonal (varimax) rotation of the extracted factors.
6) Interpretation and Validation: In ACA, interpretation and validation of results generally interact. Interpretation relies on discovering what the author clusters, factors, and map dimensions represent in terms of scholarly contributions, institutional or geographic ties, intellectual associations, and the like
沒有留言:
張貼留言