No one beginning a data science project should start from a blinking cursor.
...Templatization is a best practice for things like using common directory structure across projects...
-Megan Risdal Kaggle Product Lead.
This project contains the files and settings commonly used in analysis projects with R. A developer can start an analysis repository more quickly by copying these files. The purpose of each directory is described in its README file. Some aspects are more thoroughly described in Collaborative Data Science Practices.
The repo contains two pipelines that aim to be simple enough to understand, yet complex enough to mimic aspects frequently seen in analysis projects.
The simplest example involves a csv that is lightly groomed and saved as an rds file. A knitr Rmd file analyzes the rds; the text, graphs, and tables are saved as a self-contained html. The html file is veyr protable; it can be saved on a drive, emailed to a colleauge, or publically served on a website.
Most nontrivial data science projects require multiple sources to address a single issue. This example uses three sources: (a) longitudinal measurements for individuals across time (mlm.csv), (b) static county characteristics (county.csv), and (c) longintudinal county-level characteristics (te.csv). Each csv is independently groomed and loaded into its own database table (in db.sqlite) by an ellis lane. Conventional statistical software is not designed to digest multiple data rectangles; a scribe transforms multple database-normalized tables into a single rds that can be analyzed directly. In this case, the mlm.rds supports two analyses: a conventional report of statistical inferences intended for subject-experts concerned with complex hypotheses, and a dashboard of simplified patterns intended for administrators concerned with operational progress. The te.rds supports a comparison of the time and effort results between counties.
