Generally, numerical data should be stored in UTF-8 encoded plain text files. In order to make the data file self-descriptive, any data file must contain:
-
a header containing metadata
-
a values section containing the actual data
Guidelines for the values section:
-
Only basic 7-bit ASCII characters should be used in values section
-
Values in a row must be separated by a delimiter
-
Every row must end with a line break. Use your the default line break character of your operating system
-
The same delimiter structure and line ending must be used in all rows, also in the last row.
-
Allowed delimiters are
, ; : Space TAB
-
Preferred delimiters are
, ; :
-
Data can be structured either as
independent variable delimiter dependent var. (delimiter dep. var. etc.) line break character
or as dependent values on a grid. In the latter case, the structure of the grid must be specified in the header.
- Date-time values should be given according to ISO 8061 (e.g., YYYY-MM-DDThh:mm:ss, 2001-08-08T19:35:02). UTC should always be used (otherwise, it must be indicated in the header).
Guidelines for the header section:
- The header section must be clearly marked as such and distinguished from the values section. This can be done by marking its beginning/end, by using a special character (such as #) at the beginning of each line, or by specifying the number of header lines.
- The header section must as a minimum contain a line specifying the quantities given in the values section and their respective units, including the delimiters, plus information about author(s) of the file and a link to the license agreement.
- The header should contain: information about how, when and where the data have been acquired and processed, instruments used and their setup (basically, all information needed if the reader wanted to reproduce the measurement). A contact of one of the authors might also be useful.
- Error estimates can either be included in the header or the values section
- It is possible to link to external resources if metadata are too large for the header. In this case, DOI should be used.