Abstract
AbstractChance models of scientific creative productivity allow estimation of researcher capacity. One prominent such model is the Q model in which the impact of a scholarly work is modeled as a multiplicative function of researcher capacity and a potential impact (i.e., luck) parameter. Previous work estimated researcher capacity based on an approximation of the Q parameter. In this work, however, I outline how the Q model can be estimated within the framework of generalized linear mixed models. This way estimates of researcher capacity (and all other parameters of the Q model) are readily available and obtained by standard statistical software packages. Usage of such software further allows comparing different distributional assumptions and calculation of reliability of the Q parameter (i.e., researcher capacity). This is illustrated for a large dataset of multidisciplinary scientists (N = 20,296). The Poisson Q model was found to have negligibly better predictive accuracy than the original normal Q model. Reliability estimates revealed excellent reliability of Q estimates with conditional reliability being mostly in acceptable ranges. Reliability of Q parameter estimates further depended heavily on the number of publications of a scientist with reliability increasing with the number of papers. The future and limitations of the Q model in the context of researcher capacity estimation are thoroughly discussed.
Funder
Westfälische Wilhelms-Universität Münster
Publisher
Springer Science and Business Media LLC
Subject
Library and Information Sciences,Computer Science Applications,General Social Sciences
Reference28 articles.
1. Akaike, H. (1998). Information theory and an extension of the maximum likelihood principle. In E. Parzen, K. Tanabe, & G. Kitagawa (Eds.), Selected Papers of Hirotugu Akaike (pp. 199–213). Springer. https://doi.org/10.1007/978-1-4612-1694-0_15
2. Alvarez, P., & Pulgarín, A. (1996). Application of the Rasch model to measuring the impact of scientific journals. Publishing Research Quarterly, 12(4), 57–64. https://doi.org/10.1007/BF02680575
3. Brooks, M. E., Kristensen, K., Van Benthem, K. J., Magnusson, A., Berg, C. W., Nielsen, A., Skaug, H. J., Machler, M., & Bolker, B. M. (2017). GlmmTMB balances speed and flexibility among packages for zero-inflated generalized linear mixed modeling. The R Journal, 9(2), 378. https://doi.org/10.32614/RJ-2017-066
4. Brown, A., & Croudace, T. J. (2015). Scoring and estimating score precision using multidimensional IRT models. In S. P. Reise & D. A. Revicki (Eds.), Multivariate applications series. Handbook of item response theory modeling: Applications to typical performance assessment (pp. 307–333). Routledge/Taylor & Francis Group.
5. Commenges, D., Proust-Lima, C., Samieri, C., & Liquet, B. (2015). A universal approximate cross-validation criterion for regular risk functions. The International Journal of Biostatistics. https://doi.org/10.1515/ijb-2015-0004