ASA Connect

I'm in the middle of writing an article about peer-reviewed datasets (datasets that are included as part of a peer-reviewed publication). One of my concerns is that very few peer-reviewed datasets list the conditions and restrictions on the use of the data. For example, am I allowed to keep a copy of one of these datasets on my public github site? A cursory review through Google shows some sources that indicate that data (with a very few exceptions) is not copyrightable. Data are facts and facts are not copyrightable. Other sources mention the concept of a creative compilation that would allow datasets to be copyrighted. If anyone can provide information, ideally through a source that would be considered definitive, I would greatly appreciate it.

------------------------------
Stephen Simon, blog.pmean.com
Independent Statistical Consultant
P. Mean Consulting
------------------------------

Hi Stephen,

The researchers here might be able to point you in a useful direction: https://www.paperswithcode.com/about
(By pure coincidence I came across this shortly after reading your post.)

Best,
Glen

------------------------------
Glen Wright Colopy
DPhil Oxon
Data Scientist at Cenduit LLC, Durham, NC
------------------------------

Original Message:
Sent: 02-12-2021 22:01
From: Stephen Simon
Subject: Copyright and datasets

I'm in the middle of writing an article about peer-reviewed datasets (datasets that are included as part of a peer-reviewed publication). One of my concerns is that very few peer-reviewed datasets list the conditions and restrictions on the use of the data. For example, am I allowed to keep a copy of one of these datasets on my public github site? A cursory review through Google shows some sources that indicate that data (with a very few exceptions) is not copyrightable. Data are facts and facts are not copyrightable. Other sources mention the concept of a creative compilation that would allow datasets to be copyrighted. If anyone can provide information, ideally through a source that would be considered definitive, I would greatly appreciate it.

------------------------------
Stephen Simon, blog.pmean.com
Independent Statistical Consultant
P. Mean Consulting
------------------------------

Hello there.
Regarding, technically "copyright",  there is so much i can talk about and help. But as important or more is the ethics part of the  datasets, how deidentified they actually are and hence that part is what may preclude open sharing of some data? Different jurisdictions have different rules and protection acts for data and data sharing.

Interesting to keep an eye on this discussion.

------------------------------
Alberto Nettel-Aguirre
Professor of Biostatistics
University of Wollongong
------------------------------

Original Message:
Sent: 02-13-2021 21:30
From: Glen Colopy
Subject: Copyright and datasets

Hi Stephen,

The researchers here might be able to point you in a useful direction: https://www.paperswithcode.com/about
(By pure coincidence I came across this shortly after reading your post.)

Best,
Glen

------------------------------
Glen Wright Colopy
DPhil Oxon
Data Scientist at Cenduit LLC, Durham, NC

Original Message:
Sent: 02-12-2021 22:01
From: Stephen Simon
Subject: Copyright and datasets

I'm in the middle of writing an article about peer-reviewed datasets (datasets that are included as part of a peer-reviewed publication). One of my concerns is that very few peer-reviewed datasets list the conditions and restrictions on the use of the data. For example, am I allowed to keep a copy of one of these datasets on my public github site? A cursory review through Google shows some sources that indicate that data (with a very few exceptions) is not copyrightable. Data are facts and facts are not copyrightable. Other sources mention the concept of a creative compilation that would allow datasets to be copyrighted. If anyone can provide information, ideally through a source that would be considered definitive, I would greatly appreciate it.

------------------------------
Stephen Simon, blog.pmean.com
Independent Statistical Consultant
P. Mean Consulting
------------------------------

My thoughts are exactly along Prof. Nettel-Aguirre's lines. Since the removal of PIIs on its own is not necessarily sufficient as a deidentification practice in some emerging regulations (GDPR for example) yet different people have different understandings of what deidentification means, I am curious how the research community will handle data sets with human subjects--for example, how would you handle should a data set be deemed not completely anonymous by what turns out to be an applicable regulation?

------------------------------
Michiko Wolcott
Principal Consultant
Msight Analytics
------------------------------

Original Message:
Sent: 02-15-2021 17:03
From: Alberto Nettel-Aguirre
Subject: Copyright and datasets

Hello there.
Regarding, technically "copyright",  there is so much i can talk about and help. But as important or more is the ethics part of the  datasets, how deidentified they actually are and hence that part is what may preclude open sharing of some data? Different jurisdictions have different rules and protection acts for data and data sharing.

Interesting to keep an eye on this discussion.

------------------------------
Alberto Nettel-Aguirre
Professor of Biostatistics
University of Wollongong

Original Message:
Sent: 02-13-2021 21:30
From: Glen Colopy
Subject: Copyright and datasets

Hi Stephen,

The researchers here might be able to point you in a useful direction: https://www.paperswithcode.com/about
(By pure coincidence I came across this shortly after reading your post.)

Best,
Glen

------------------------------
Glen Wright Colopy
DPhil Oxon
Data Scientist at Cenduit LLC, Durham, NC

Original Message:
Sent: 02-12-2021 22:01
From: Stephen Simon
Subject: Copyright and datasets

I'm in the middle of writing an article about peer-reviewed datasets (datasets that are included as part of a peer-reviewed publication). One of my concerns is that very few peer-reviewed datasets list the conditions and restrictions on the use of the data. For example, am I allowed to keep a copy of one of these datasets on my public github site? A cursory review through Google shows some sources that indicate that data (with a very few exceptions) is not copyrightable. Data are facts and facts are not copyrightable. Other sources mention the concept of a creative compilation that would allow datasets to be copyrighted. If anyone can provide information, ideally through a source that would be considered definitive, I would greatly appreciate it.

------------------------------
Stephen Simon, blog.pmean.com
Independent Statistical Consultant
P. Mean Consulting
------------------------------

I am Data Editor for the American Economic Association, and face this issue on a regular basis. Good guidance is provided by Stodden (2009) https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1362040. Facts cannot be copyrighted (in the US!), but products derived from those facts typically do allow to be copyrighted (i.e., the presentation, organization, etc.). In general, for the data included in replication packages posted on our repository, we require that authors provide evidence that they have re-distribution rights, whether by explicit permission of the original data owner, or via some license (e.g., CC-BY or ODBL). Once the replication package is posted, it is subject to the license shown on the specific deposit (typically CC-BY, but where there is a license file, it applies). This can sometimes be tricky, and has only been (imperfectly, probably) monitored since July 2019 - any deposits from before that may not be quite as informative. For a discussion of some of the challenges, see my report from 2019 (https://pubs.aeaweb.org/doi/pdfplus/10.1257/pandp.109.718).

From a practical point of view, I would never suggest posting a dataset unless you have permission from the original authors (of the data, not necessary of the original article), or the presence of an obvious license. Do not assume that the authors of the posted dataset have done their homework - if you repost data that is subject to re-distribution restrictions by the original data owner, you probably won't prevail by pointing at the previous author by saying "he also did it". We have had multiple instances of that when the previous replication package clearly infringed on data use agreements.


------------------------------
Lars Vilhuber
Managing Editor, Journal of Privacy and Confidentiality
Data Editor, American Economic Association
Economist
Cornell University - Labor Dynamics Institute
------------------------------

Original Message:
Sent: 02-12-2021 22:01
From: Stephen Simon
Subject: Copyright and datasets

I'm in the middle of writing an article about peer-reviewed datasets (datasets that are included as part of a peer-reviewed publication). One of my concerns is that very few peer-reviewed datasets list the conditions and restrictions on the use of the data. For example, am I allowed to keep a copy of one of these datasets on my public github site? A cursory review through Google shows some sources that indicate that data (with a very few exceptions) is not copyrightable. Data are facts and facts are not copyrightable. Other sources mention the concept of a creative compilation that would allow datasets to be copyrighted. If anyone can provide information, ideally through a source that would be considered definitive, I would greatly appreciate it.

------------------------------
Stephen Simon, blog.pmean.com
Independent Statistical Consultant
P. Mean Consulting
------------------------------

If data is already posted, why would we need to repost it? Wouldn't a statement like, "The data we used is available here. <enter website address>"

Also, what if you are reusing data for publication in the same journal? Or, what if you want to reanalyze "textbook" data, as in a data set that comes from a textbook you used? How does that effect things?

------------------------------
Andrew Ekstrom

Statistician, Chemist, HPC Abuser;-)
------------------------------

Original Message:
Sent: 02-15-2021 10:09
From: Lars Vilhuber
Subject: Copyright and datasets

I am Data Editor for the American Economic Association, and face this issue on a regular basis. Good guidance is provided by Stodden (2009) https://papers.ssrn.com/sol3/papers.cfm?abstract_id=1362040. Facts cannot be copyrighted (in the US!), but products derived from those facts typically do allow to be copyrighted (i.e., the presentation, organization, etc.). In general, for the data included in replication packages posted on our repository, we require that authors provide evidence that they have re-distribution rights, whether by explicit permission of the original data owner, or via some license (e.g., CC-BY or ODBL). Once the replication package is posted, it is subject to the license shown on the specific deposit (typically CC-BY, but where there is a license file, it applies). This can sometimes be tricky, and has only been (imperfectly, probably) monitored since July 2019 - any deposits from before that may not be quite as informative. For a discussion of some of the challenges, see my report from 2019 (https://pubs.aeaweb.org/doi/pdfplus/10.1257/pandp.109.718).

From a practical point of view, I would never suggest posting a dataset unless you have permission from the original authors (of the data, not necessary of the original article), or the presence of an obvious license. Do not assume that the authors of the posted dataset have done their homework - if you repost data that is subject to re-distribution restrictions by the original data owner, you probably won't prevail by pointing at the previous author by saying "he also did it". We have had multiple instances of that when the previous replication package clearly infringed on data use agreements.


------------------------------
Lars Vilhuber
Managing Editor, Journal of Privacy and Confidentiality
Data Editor, American Economic Association
Economist
Cornell University - Labor Dynamics Institute

Original Message:
Sent: 02-12-2021 22:01
From: Stephen Simon
Subject: Copyright and datasets

I'm in the middle of writing an article about peer-reviewed datasets (datasets that are included as part of a peer-reviewed publication). One of my concerns is that very few peer-reviewed datasets list the conditions and restrictions on the use of the data. For example, am I allowed to keep a copy of one of these datasets on my public github site? A cursory review through Google shows some sources that indicate that data (with a very few exceptions) is not copyrightable. Data are facts and facts are not copyrightable. Other sources mention the concept of a creative compilation that would allow datasets to be copyrighted. If anyone can provide information, ideally through a source that would be considered definitive, I would greatly appreciate it.

------------------------------
Stephen Simon, blog.pmean.com
Independent Statistical Consultant
P. Mean Consulting
------------------------------