The publication records are periodically updated following updates from the Web of Science®. The last update to this site occurred on February 27, 2017.
Filtering and disambiguation
One challenge to studying intra-school collaboration by using publication data is in the proper association of papers with schools, departments, and faculty. This section describes the filtering process used to associate publication data with the university, schools within the university, departments within the schools, and individual faculty members.
Papers are associated with Northwestern University by using the organization tag of each paper. Papers that have at least one affiliation with an organization of NORTHWESTERN UNIV are included.
School and Department level
Among the set of papers that are associated with the university, papers are associated with departments by matching the suborganization tag with the department or school. For example, a paper that has an affiliation with an organization of NORTHWESTERN UNIV and a suborganization of DEPT MECH ENG would be associated with the Department of Mechanical Engineering. After a paper is associated with its departments, it is associated with its schools according to the school to which the department belongs. In the example, this paper would be associated with the McCormick School of Engineering. Note that only certain predetermined departments and schools are included in this analysis.
The association method for faculty members is identical to the department level association. Faculty members are associated with papers by matching their “faculty roster name” (e.g. Jane Doe) with an author name in the Web of Science® given by their last name and first initial(s) (e.g. “DOE, J”). To increase accuracy, manual curation of the Web of Science® names that are associated with a faculty roster name was performed and compared with publication numbers that were listed on CVs that were available on the web. Note that for researchers that have a more common name, there is an increased likelihood of falsely associating a paper with them that they did not actually author (false positive). Note that, due to the higher false positive rate at the faculty level, the association of papers to individual faculty members is done independently of the school and department level association mentioned above. Please note that only select faculty are included, namely all McCormick faculty and some Weinberg faculty who have coauthored papers with McCormick faculty from 2002-2017.
Sources of error
As with any analysis using empirical data, it is important to understand the sources and types of errors that can occur as discussed below.
False positives occur when a paper that was not actually authored by a school, department, or faculty member is attributed to them. This usually occurs at the faculty level when, for instance, there is a different author with the same name on a paper or when there is a misspelling of the actual author who coincides with a different author.
False negatives occur when a paper that was authored by a school, department, or faculty member is not attributed to them. This can occur at the faculty level when an author publishes under a name that is not in the list of author names or at the department or school level when an author publishes under an affiliation that is not listed in the list of department names. Another source of false negatives is if a faculty member publishes in journals that are not covered in the Web of Science®. Note that the coverage of the Web of Science® varies by discipline and is constantly changing; one should carefully account for the specific publication landscape when comparing faculty across disciplines.
The following describes the metrics used to scale the circle radius and line width in the network diagrams.
- Number of papers: The absolute number of publications that are associated with the school, department, or faculty member in the highlighted year.
- Paper per author: The number of papers per distinct author name for that school, department, or faculty member. For example, if a department publishes 10 papers in 2005 and there are 4 distinct author names that are associated with all 10 papers (e.g. “J DOE”, “JQ PUBLIC”, “JT PLUMBER”, “J SIXPACK”), then the papers per author is 2.5=10/4.
- Number of collaborations: The number of published papers that are coauthored by two schools, departments, or faculty members.
- Collaboration density: The fraction of papers published by two entities (schools, departments, or faculty members) that are collaborations between those two entities.
At the school level, the McCormick School of Engineering is placed at the center of the diagram, and the other schools are arranged radially. Node placement does not change between years, allowing a user to track a single school easily over time.
At the department level, the departments in the McCormick School of engineering are arranged radially in the center of the diagram, and the other departments are arranged radially outside of the central circle. As in the school level diagrams, the node placement does not change between years, allowing a user to track a single department easily over time.
At the faculty level, the nodes are placed using a force-directed algorithm. In contrast to the school-level and department-level diagrams, the nodes change position each year. While this feature makes it more difficult to track an individual faculty member over time, it makes it easier to see who a faculty member collaborates with in a single year.
Faculty members with joint appointments in two or more departments often include more than one affiliation on publications. At the department and school level, these papers with multiple affilitations will be counted as collaborations (even if there is only a single author).
All data is available for download in CSV format. This includes the annual number of papers and papers per author for each school, department, and faculty member as well as the annual number of collaborations for McCormick, McCormick's departments, and McCormick's faculty members.