I recently (Gow, 2026) proposed an extension to the data science “whole game” of R for Data Science (Wickham et al., 2023). In Gow (2026), I used Australian stock price data to illustrate the data curation process and, in this note, I use US bank “Call Report” data as a second illustration. In effect, I provide complete instructions for building a high-performance data library covering all Call Reports data provided by the FFIEC Bulk Data website that can be construction in less than ten minutes on fast hardware (or a couple of hours on an older machine). I also give a few brief demonstrations of how to use the curated data, with examples for both R and Python. I conclude by discussing challenges encountered during processing and offering some observations about AI and data curation.

In this note, I explore some recent changes in the open-source R package dbplyr to illustrate some of the beauty of how open-source software evolves in practice. In particular, I offer two case studies where features requested by users became reality in dbplyr, which may be my favourite R package.

Plot of feasible return parameters with portfolios of two risky assets

Plot of feasible portfolio returns given two risky assets by correlation

R code to replicate examples from [here](https://wrds-web.wharton.upenn.edu/wrds/support/Accessing%20and%20Manipulating%20the%20Data/_007R%20Programming/_001Using%20R%20with%20WRDS.cfm).

House prices over time

Test