Courrier des statistiques N7 - 2022
Some good software development techniques for the self-study statistician (or "How to count, how to code")
In addition to skills in statistical methodology and a good knowledge of available data sources, the profession of statistician requires to be comfortable with IT tools. Programs not only produce the statistical results, they can also become deliverables, either as evidence or as reusable tools for future work. Having this in mind, the statistician must acquire software development best practices that will allow him to guarantee an easy understanding of his programs by other users, or re-appropriation by himself in future developments.
These best practices cover all aspects of the software development cycle: definition of requirements, program architecture, programming styles, technical choices, development tools and testing. They will make it possible to guarantee an appropriate answer to users' needs, after a step of questioning these needs, and quality of the results obtained while controlling the development costs. And more important, by making the programs easily readable, they will help the producer of official statistics to communicate on his methodological choices and how to use the data, and thus strengthen the confidence of his users.
Paru le :19/02/2024