Ranking Algorithm Review
\(\varepsilon\) Ranking Algorithm Review:
The ε-index uses a researcher's citation data from Google Scholar, requiring key pieces of information: citations of the top-cited paper, the i10-index, the h-index, and the year of the first peer-reviewed publication.
The model involves calculating a power-law-like relationship between citation frequencies (\(i_{10}, h\)) and values (\(10, h, \text{ and } cm\)), then applying a linear model (\(y \sim \alpha + \beta x\)) where \(y\) is the citation frequency and \(x\) is the citation value. The ε-index is derived as the residual from this model, offering a relative metric of citation performance.
Ranking Algorithm Review
- Ranking algorithm used to establish a relationship between citation frequencies and their corresponding values, which is then indexed, smoothed, and standardized.
- The algorithm works by a simple linear model \(y = \alpha + \beta_x\) where
\[
y = \log_e = \left| \begin{array}{c}
i_{10} \\
h \\
1 \end{array} \right|
\]
and
\[
x = \log_e = \left| \begin{array}{c}
10 \\
h \\
c_{m} \end{array} \right|
\]
- Where (\(h\) is both a frequency (\(y\)-axis) and a value (\(x\)-axis).
- The data are then standardized to index \(A_{rel}\) - Citation mass which then becomes the dependent variable in the linear model
\(A_{rel} = \lambda + \Theta \log_e t \)
- The residual \((\varepsilon)\) \(A_{Rel}\) then gets computed as a measure of citation performance relative to the sample data.
- The residual \((\varepsilon)\) then gets ranked highest to lowest such that positive values of \((\varepsilon)\) are above "expected performance" and negative values are below "expected performance".
Note: Data Pulled from Google Scholar December 2023. Special thanks to Alex Strong for his contribution to gathering this data.
Note: Faculty without Google Scholar profiles associated with the University of Arizona are not included.
Note: Questions, Comments, and Concerns can be directed to ALVSCE Data Solutions.