DataSHIELD disclosure settings - migration of pages to new wiki

Statistical development
, ,
1

I have migrated the disclosure settings and table of dsBase checks pages from the old wiki to the new one. On the old wiki, pages that are migrated will contain an information banner detailing the location of the new page on the wiki - to mark pages that are migrated.

The page sits under statistical development theme - i have created a contents subheading on the landing page of the stats theme. A direct link to the page is here:

https://wiki.datashield.org/en/statdev/disclosure-checks

Please note this is dsBase functions correct as of 2022. I know that new functions have been added, if someone in stats theme is able to update either table, please go ahead. Also if you ave developed a DataSHIELD package that makes use of any of the disclosure settings listed in the table, please add your server side package to the list and compile your own table of functions listing which threshold checks are done.

2

Hi Becca, all

*** this is not an attack on Datashield - merely a plea for transparency****

The wiki entry is factually incorrect when it says:

Disclosure checks are mapped to current best practice for output checking Welpton, Richard (2019): SDC Handbook

In fact DataShield differs in various ways from current norms and recommendations of best practice.

  • To give one example , because of the way that threshold checks are done, it allows tables to be returned which have zero entries in cells, which in turns poses large risks of what is known in the SDC field as ‘class disclosure’.
3

Im just adding here the additional text from Jim’s message which included an example to help readers. This has been truncated from the post somehow.

“e.g. Datashield would release a table of diabetes diagnosis (type1,type2, none) vs some protected characteristic (e.g. ethnic group). With 0 in the ’none’ column for a given group — which discloses the private information that all people in that group had diabetes.”

I will close comment on this duplicate post as the discussion about this topic continues in this post in the Support forum. Could DataShield be adapted to meet accepted SDC norms?