Computational Social Science
My work is substantive at its core. The methods I use to investigate my substantive topics are highly dependent on computational social science (CSS).
Reproducible Repositories
Every study I conducted and published, including roughly all working papers, include a reproducible repository. These take on different formats, that follow the development of my CSS approaches and knowledge over time.
Early studies used Word or PDF documents to summarize the code and workflow.
2010. Economic Equality - this paper originally was published with a Google Drive link to a single PDF containing the entire workflow in Stata and Mplus.
2011. A Clash of Civilizations - likewise this paper has a single PDF for users to follow to replicate the study in Stata.
Later studies used OSF repositories that included code, data and technical files.
The use of Github and the transition to R opened up new possibilities.
Two examples of reproducible repositories
OSF Repo for 'No Generalizable Effect' (2019) with Stata code. Complete with Wiki showing how to reproduce our tables and figures and extra analyses.
GitHub Repo for 'A Hidden Universe' (2023) with various code embedded in R.
Apps and Visualizations
The social sciences I experienced early on, had simple black and white scatterplots on a good day, and mostly just tables otherwise. It is not that the technology was not there, more that visualization was nothing that was taught in the social sciences. This is changing, and I think it brings great improvements to the understandability of our findings.
For example, my 2019 paper with Carola Hommerich could have simply reported a grand mean slope and that it was not significantly different from zero because the dispersion of within country slopes was so great. But in my opinion, the following graphic does more than words.
I could have also reported a huge table of regression results in our 'Hidden Universe' paper, but the following seemed to do the job much better
The visualization above changes depending on model specifications. This is not really visualizable in a static figure. Learning to program an interactive application solved this problem, and provides a powerful tool for users to explore our results even if they have no knowledge of the methods or programming language.