From the about page:
This website shows how Federal agencies are performing on the latest Open Data Policy (M-13-13) using the guidance provided by Project Open Data. It also provides many other other tools and resources to help agencies and other interested parties implement their open data programs. Features include:
- A dashboard to track the progress of agencies implementing Project Open Data on a quarterly basis
- Automated analysis of URLs provided within metadata to see if the links work as expected
- A validator for v1.0 and v1.1 of the Project Open Data Metadata Schema
- A converter to transform CSV files into JSON as defined by the Project Open Data Metadata Schema Link broken as of 4 January 2014. Site notified.
- An export API to export from the CKAN API and transform the metadata into JSON as defined by the Project Open Data Metadata Schema
- A changeset viewer to compare a data.json file to the metadata currently available in CKAN (eg catalog.data.gov)
You can learn more by reading the main documentation page.
The main documentation defines the “Number of Datasets” on the dashboard as:
This element accounts for the total number of all datasets listed in the Enterprise Data Inventory. This includes those marked as “Public”, “Non-Public” and “Restricted”.
If you compare the “Milestone – May 31st 2014” to November, the number of data sets increases in most cases, as you would expect. However, both the Department of Commerce and the Department of Health and Human Services, had decreases in the number of available data sets.
On May 31st, the Department of Commerce listed 20488 data sets but on November 30th, only 372. A decrease of more than 20,000 data sets.
On May 31st, the Department of Health and Human Services listed 1507 data sets but on November 30th, only 1064, a decrease of 443 data sets.
Looking further, the sudden decrease for both agencies occurred between Milestone 3 and Milestone 4 (August 31st 2014).
Sounds exciting! Yes?
Yes, but this illustrates why you should “drill down” in data whenever possible. And if not possible in interface, check other sources.
I followed the Department of Commerce link (the first column on the left) to the details of the crawl and thence the data link to determine the number of publicly available data sets.
As of today, 04 January 2014, the Department of Commerce has 23,181 datasets and not the 372 reported for Milestones 5 or the 268 reported for Milestone 4.
As of today, 04 January 2014, the Department of Health and Human Services has 1,672 datasets and not the 1064 reported for Milestones 5 or the 1088 reported for Milestone 4.
The reason(s) for the differences are unclear and the dashboard itself offers no explanation for the disparate figures. I suspect there is some glitch in the automatic harvesting of the information and/or in the representation of those results in the dashboard.
Always remember that just because a representation* claims some “fact,” that doesn’t necessarily make it so.
*Representation: Bear in mind that anything you see on a computer screen is a “representation.” There isn’t anything in data storage that has any resemblance to what you see on the screen. Choices have been made out of your sight as to how information will be represented to you.
As I mentioned yesterday, there is a common and naive assumption that data as represented to us has a reliable correspondence with data held in storage. And that the data held in storage has a reliable correspondence to data as entered or obtained from other sources.
Those assumptions aren’t unreasonable, at least until they are. Can you think of ways to illustrate those principles? I ask because at least one way to illustrate those principles makes an excellent case for open source software. More on that anon.