The YODA principles are a small set of guidelines that can make a huge difference towards reproducibility, comprehensibility, and transparency in a data analysis project. By applying them in your own midterm analysis project, you have experienced their immediate benefits.
You also noticed that these standards are not complex – quite the opposite, they are very intuitive. They structure essential components of a data analysis project – data, code, potentially computational environments, and lastly also the results – in a modular and practical way, and use basic principles and commands of DataLad you are already familiar with.
There are many advantages to this organization of contents.
Having input data as independent dataset(s) that are not influenced (only consumed) by an analysis allows for a modular reuse of pure data datasets, and does not conflate the data of an analysis with the results or the code. You have experienced this with the
Keeping code within an independent, version-controlled directory, but as a part of the analysis dataset, makes sharing code easy and transparent, and helps to keep directories neat and organized. Moreover, with the data as subdatasets, data and code can be automatically shared together. By complying to this principle, you were able to submit both code and data in a single superdataset.
Keeping an analysis dataset fully self-contained with relative instead of absolute paths in scripts is critical to ensure that an analysis reproduces easily on a different computer.
DataLad’s Python API makes all of DataLad’s functionality available in Python, either as standalone functions that are exposed via
datalad.api, or as methods of the
Datasetclass. This provides an alternative to the command line, but it also opens up the possibility of performing DataLad commands directly inside of scripts.
Including the computational environment into an analysis dataset encapsulates software and software versions, and thus prevents re-computation failures (or sudden differences in the results) once software is updated, and software conflicts arising on different machines than the one the analysis was originally conducted on. You have not yet experienced how to do this first-hand, but you will in a later section.
Having all of these components as part of a DataLad dataset allows version controlling all pieces within the analysis regardless of their size, and generates provenance for everything, especially if you make use of the tools that DataLad provides. This way, anyone can understand and even reproduce your analysis without much knowledge about your project.
The yoda procedure is a good starting point to build your next data analysis project up on.
6.4.1. Now what can I do with it?¶
Using tools that DataLad provides you are able to make the most out of your data analysis project. The YODA principles are a guide to accompany you on your path to reproducibility and provenance-tracking.
What should have become clear in this section is that you are already equipped with enough DataLad tools and knowledge that complying to these standards felt completely natural and effortless in your midterm analysis project.