All data sets published on DataShare must be accompanied by documentation. At minimum, your data documentation should address:
The simplest and most flexible way to document your data is a README file. README files are human-readable text files usually saved as plain text (.txt) or as a PDF. It's possible to insert a codebook or data dictionary into a README but it may not always be practical.
The terms codebook and data dictionary are often interchangeable. In general codebooks are more formal in content and format, often only addressing data variables, while data dictionaries may incorporate elements of a README file and list sources, methodologies, etc.
Data dictionaries and codebooks work best for tabular data (i.e. spreadsheets and tables). They explain the attributes of the data elements that are not within the data itself, such as what each column of the data represents and how variables are defined. For example, a column titled "date" does not tell you why the date matters, only that it is one.