Principles of Reproducible Research

Insight Statistics adheres to principles of reproducible research. This helps ensure scientific integrity and increase efficiency.  

With reproducible research, we add code to research publications and reports. The code is used to generate results, and then these are automatically inserted in the correct location within the text. Although it's possible to use larger chunks of code, we generally prefer to insert a single line of code that references a program. The single line of code is placed where you want a table or figure to appear. Whereas the program that makes the table or figure might be quite complex, the line of code that references it is quite simple and easy to understand. In fact, it's not much more complicated than writing "Insert table here" or "Insert figure here". A click of a button compiles the research publication or report. The code for creating each table or figure is run, each table or figure is inserted in the correct location, and a research publication or report is output. Formatting instructions are specified beforehand. So the formatting of the research publication or report is automatic.

Reproducible research dynamically links software, code, data, analysis, output, and scientific reporting. This makes the research process self-documenting. It creates a record of exactly what software environment was used to generate the results, what code was used, what data were used, how the data were analyzed, and how the outputs were produced. Tools for reproducible research also can be used to document the thought process that took place and the changes that were made as the research unfolded. When combined with clearly-written and well-documented code, a process of reproducible research makes it easy to replicate and verify our findings. It also enhances efficiency by providing a clear structure and set of processes and by making it much easier and quicker to change the analysis or update the analysis using new data.

Insight Statistics uses several tools to conduct reproducible research. These include R, RStudio, markup languages, and Unix-like shell programs. An R package called knitr is used to connect R-based analyses to high-quality presentation documents. These documents are generated in Portable Document Format (PDF) using markup languages such as LaTeX and Markdown. A similar process can be used to generate documents using word processing software like Microsoft Word.