data planning / what is a dataset

[k-rns]

Gitbook: https://app.gitbook.com/o/-LiioKfZ_6hmmUkbozT5/s/-MQxi3LV7ImgXRCb9ZcM/data-planning/what-is-a-dataset

[k-rns]

Comments on first draft (adjusted from internal documentation)

[Shannon Rauch] [Aug 29th at 13:15]
Re: "What is a Dataset?" page:
I wonder if we should link to an example landing page for each data type?
Just sort of a usability thing - there are a lot of sentences that say 'such and such can be found here' with the 'here' being a clickable link to another page. If those links ever break for whatever reason, it might be hard to figure out where they're supposed to be going. Instead of "Details on the supported format and organization of tabular data can be found here." for example, it might be better to say "See the 'Organizing Data Tables' page for more detail on the supported format and organization of tabular data."

[Danie Kinkade ]
Re: "What is a Dataset?" page: (with apologies for length of comments!)
In the very first line, we call a a dataset a "logical collection of information". But, I think we may want to clarify that we understand a model or piece of code is not the same as a table of observations or collection of images, even though we represent them equivalently on our data pages. I think a "logical collection of information" is tough to distinguish from metadata. And although I realize the line b/t the two can be blurry, starting out saying data is information seems vague and possibly contradictory to me.
The OSTP has just updated its 2013 PARR memo to strengthen the immediate public release of research scholarly publication/information and... data. Maybe framing our "what is a dataset" intro with the US OSTP definition data could help lead into the content of this page.
Would something like this be helpful or more confusing?:
As a federally-funded data repository, BCO-DMO aligns with the U.S. Office of Science and Technology Policy (OSTP) concept of data as outlined in its 2022 Memorandum "Ensuring Free, Immediate, and Equitable Access to Federally Funded Research" where: “scientific data” include the recorded factual material commonly accepted in the scientific community as of sufficient quality to validate and replicate research findings. Such scientific data do not include laboratory notebooks, preliminary analyses, case report forms, drafts of scientific papers, plans for future research, peer-reviews, communications with colleagues, or physical objects and materials, such as laboratory specimens, artifacts, or field notes.
Also, in the first paragraph, I notice that we mix the concept of a table and file together and we may want to be explicit about the distinction. So maybe change:
"A BCO-DMO dataset usually contains one data table (example formats: csv, Excel, tsv) or a collection of files that contain the same type of data, like a collection of model output files in .netcdf format or a collection of coral quadrat photos." to
"A BCO-DMO dataset usually consists of a file containing a single data table (example formats: csv, Excel, tsv) or a collection of files that contain the same type of data, like a collection of model output files in .netcdf format or a collection of coral quadrat photos."
I agree with Shannon on providing links to representative datasets, not only for the data types but for the examples we list above that section.
For the last section on metadata:
I think the statement "Each dataset needs ... (information describing the data)" could be edited to "Each dataset needs... (information providing sufficient context for peer review and reuse of the data)"
and suggest editing "Therefore, each dataset needs to have its own metadata page, specifically describing how the data were generated." to "Therefore, each dataset needs to have its own metadata page, describing the who/what/where/how information about the data." Since our metadata pages contain much more than simply methods.
I'd also recommend removing the sentence "Gathering the necessary metadata for a specific dataset can feel cumbersome, but a robust data management plan at the start of the project can help alleviate the burden." I'm not sure we want to draw attention and acknowledge any negative attributes to constructing metadata. It's also a big jump for the novice submitter to make the connection between a DMP and submitting metadata to us when they deposit their data, potentially years after a proposal, and assuming they even saw the DMP (if it's a student or post-doc). If we want to spin the tracking of metadata as a positive, I'd maybe allude to documenting important aspects of research data as it's collected and analyzed making it easier to provide structured metadata to us. (edited)

Dana Gerlach
09:19
I linked to the "Metadata for Datasets" page at the bottom of the "What is a Dataset" page. A few suggestions:
First paragraph, 1st sentence: "Each dataset at BCO-DMO needs to be accompanied by metadata (information describing the data) in order for a user others to understand and effectively reuse the data."
First paragraph, 3rd sentence: "Below is a are descriptions of the necessary metadata..."
Under People, do we want to mention getting ORCIDs for contributors?
Under Abstract, we should adjust to mirror what we decided on in the meeting yesterday (https://docs.google.com/document/d/1sh13WZDldBUerYiFm-TPiDgp5cCU-n6qAKwhrxs1Pek/edit#)

adyork
1 day ago
Would it be too confusing to make another addition?
...collection of files that contain the same structure and type of data.
responding to this suggested change:

My initial reaction to the suggested change of "a file containing a single data table" was that it may be confusing for them to know what to submit to us, and that we can combine like-tables into one table.
Or maybe instead of the above suggestion to change to "structure and type" we can add a little "tip" box under this statement like:
A BCO-DMO dataset usually consists of a file containing a single data table (example formats: csv, Excel, tsv) or a collection of files that contain the same type of data, like a collection of model output files in .netcdf format or a collection of coral quadrat photos.