Metadata is important for documenting your dataset so others can follow the details of your method, sources, and analysis. Good metadata will also help you more effectively manage your own project. Key pieces of metadata can include:
If your discipline has a recommended metadata schema, it can serve as a useful, standardized template for describing your data. Otherwise, it's recommended that you use README files to document your dataset.
Select a metadata standard that is most popular and used by experts in your field. These links may be useful for getting started but they are not exhaustive.
Controlled vocabularies are preferred terms that you can use to describe your dataset. Controlled vocabularies help others to find your datasets but also facilitate meta-analysis within datasets and interoperability of multiple datasets in repositories.
Some disciplines have well-established controlled vocabularies while others do not. You can look for your discipline in this directory of metadata vocabularies, or see the select examples below.
It is always preferable to use a metadata standard and/or controlled vocabulary to describe your data, but if these are not available, README files are a good alternative. Good practices for README files include (see the Cornell University Guide to Writing "Readme" Style Metadata for more information):