A dataset can keep its familiar title while its files change. If someone wants to reproduce your calculation, a link to the general project may not identify the exact material you used.
Zenodo’s versioning guidance distinguishes identifiers for particular versions from the identifier covering the work across versions. That distinction is useful when a correction or an added file could change a reported result.
For an analysis note, keep the version-specific identifier, the filenames used and the date of retrieval. Also record any selection you made inside the deposit. Downloading a collection does not mean every file contributed to the calculation.
Consider an invented release containing 200 observations. A later version corrects three records and adds twenty more. A reader using the newer files might calculate a different average even if your code is unchanged. Naming the version turns that discrepancy into an understandable comparison.
Keep the update decision explicit
If a corrected version appears, decide whether to rerun the analysis and explain that choice. Do not silently replace the input while leaving an old result and citation in place. Where the correction affects your conclusion, describe the effect rather than merely changing the link.
Repository versioning systems differ, so follow the chosen repository’s instructions instead of inventing a suffix for an identifier. A compact citation with the right version can do more for reproducibility than a long methods paragraph that never identifies its input.
Illustration selected from this site’s existing collection.