**Exploratory Analysis** ------------------------ 1. Construct the probability distribution function of the year of first publication of faculty 2. Construct the probability distribution function of the total citations of faculty 3. Cluster the subject areas of faculty using the Louvain algorithm (k=5) 4. Generate a bar plot of the number of publications per subject area per continent pre-2014, post-2014, and their difference 5. Generate a bar plot of the number of publications per CIP category per continent pre-2014, post-2014, and their difference **Complete this first stage of analysis and submit your statistical report by March 6, 2020.** ---------- **Statistical Modeling** ------------------------ 1. Construct the author networks for USA, Europe, and Australasia. In these networks, each node is a scholar and edges connecting scholars represent joint publications. Use [Gephi][1] for this construction. Scholar nodes, depending on their departmental CIP codes would be colored differently. Consider aggregating the numerous CIP codes in the dataset in three disciplinary clusters: *Biological Sciences* (green color), *Medical Sciences* (orange color), *Engineering* (magenta color). Compute the Page Rank for each scholar. 2. Construct three linear models, one for each geographic area - USA, Europe, and Australasia. The key predictor in each model would be scholar cross-disciplinarity. The response variable would be citation impact. Hence, the main goal of these models would be to investigate whether performing cross-disciplinary research in brain science correlates with impactful careers or not. If it does, then the brain science community is on a good path, because promise of success would attract more talent in cross-disciplinary research, which will eventually make the difference. If not, then new science policies would need to be instituted to change the situation. **Hint 1** ---------- Eliminate from consideration scholars who have less than 10 publications. In all likelihood, these are either fresh out of their PhD faculty or research inactive scholars. Accordingly, they are on the fringes of the academic system and can be removed without substantial loss of information. **Hint 2** ---------- While the key predictor variable is cross-disciplinarity, you would also need to add control factors to your model to account for the effects of covariates. Time is certainly a factor that you should consider, as younger scholars did not have as much time as older scholars to accrue impact. You should also consider accounting for Page Rank, as a scholar's prominent position in the academic network is likely to act as a career booster. **Hint 3** ---------- Adopt the following definition of cross-disciplinarity: A scholar from one of the three diciplinary clusters (*Biological Sciences*, *Medical Sciences*, or *Engineering*) who published at least one paper with a scholar from another cluster. Such a paper is defined as a cross-disciplinary paper. Furthermore, we strongly encourage you to perform sensitivity analysis, by re-running the models with a stricter definition of cross-disciplinarity, i.e., classifying scholars to be cross-disciplinary only when featuring more than four cross-disciplinary papers in their record. Can you figure out why this would be a good idea? **Complete this second and final stage of analysis and submit your statistical report by April 24, 2020.** [1]: https://gephi.org