Data sharing For better science

Mercè Crosas, Institute for Quantitative Social Science, Harvard University
@mercecrosas

max Planck institute for Radioastronomy, September 12, 2017

The Dataverse ProJect

This Talk

Importance of Data Sharing
- Reproducibility to verify science
- Reuse to advance science and evidence-based policy
Enabling Data Sharing
- Data Policies from journals and funding agencies
- Data Citation to find datasets, give credit to data authors
- Data Repositories as publishers of data

Data Sharing,
DATA PUBLISHING

Data sharing is "the release of research data, associated metadata, accompanying documentation, and software code for re-use and analysis in such a manner that they can be discovered on the Web and referred to in a unique and persistent way."

Data Publishing Group, 201 5

Nullius In Verba

"take NoBoDY's WORD For IT"

(motto of the Royal Society, founded in 1660,

launched first scientific journal in 1665)

Since the Beginning of Modern Science ...

University of California Curation Center, DataPub blog, August 2017

Reproducibility and Replication
(by the National Science Foundation):

The ability of a researcher to duplicate the results of a prior study

... using the same materials and procedures used by the original investigator. (reproducibility)

... if the same procedures are followed but new data are collected. (replication)

Empirical, Computational, and Statistical Reproducibility (Stodden, 2014):

Empirical: data and collection details are made freely available

Computational: code, software, hardware and implementations details are provided

Statistical: details on choice of statistics tests, model parameters are provided

Reproducibility

Reproducibility Crisis?

Trust, but verify

6 (11%) out of 53 landmark cancer biology studies could be reproduced.

39 out of 100 psychology studies could be reproduced.

Washington Post, Joel Achenbach, August 28, 2015

Nature, 2016, "1,500 scientists lift the lid on reproducibility", vol 533, Issue 734

Nature's survey of 1,576 researchers:

703 Biology

106 Chemistry

95 Earth and Environmental

203 Medicine

236 Physics and Engineering

233 Other

"In the Wf4Ever project we propose to improve the quality of science with metrics based on reproducibility and reuse, preserving decomposable thoroughly curated digital artefacts that enhances reproducibility and visibility of the experiment, as well as allowing more accurate mechanisms for credit attribution."

ShaRing Data, Code, and workflows Facilitates reproducibility and Reduces Duplication

BUT DATA SHARING is more than Posting your data in your website

External links in all articles published between 1997 and 2008 in the four main astronomy journals published by the American Astronomical Society.

70% of Links to personal websites from articles published in 1997 are Broken

how CAN we Improve

data sharing?

New Norms
New Incentives
New Technology

Castro, Crosas, Garnett, Sheridan, Altman, 2017, Journal of Scholarly Publishing

FORMAL DATA-SHARING POLICIES ARE APPLIED IN JOURNALS ACROSS DISCIPLINEs

MANY Funders require data sharing & Open data

PRIVATE RESEARCH FUNDERS

Bill and Melinda Gates Foundation Information Sharing Approach
Sloan Foundation Data Sharing Policy
Wellcome Trust Data Sharing Policy
Arnold Foundation
Moore Foundation
Robert Wood Johnson Foundation
HHMI Policy on the Sharing of Publication-Related Materials, Data and Software

PUBLIC RESEARCH FUNDERS

Department of Agriculture
Department of Commerce
Department of Defense
Department of Education
Department of Energy
Department of Health and Human Services
- Agency for Healthcare Research and Quality (AHRQ)
- Assistant Secretary for Preparedness and Response (ASPR)
- Center for Disease Control and Prevention (CDC)
- Food and Drug Administration (FDA)
- National Institutes of Health (NIH)
Department of Homeland Security
Department of Housing and Urban Development
Department of Interior
Department of Labor
Department of Transportation
Department of Veterans Affairs
Environmental Protection Agency (EPA)

" We believe that both as a matter of fairness and as a matter of providing an incentive for data sharing, the persons who initially gathered the data should receive appropriate and standardized credit that can be used for academic advancement, for grant applications, and in broader situations."

From 10,555 studies with gene expression microarray data:

Studies that shared data received 9% more citations

Data reuse by other researchers continued for 6 years

Data sharing increases citations

Piwowar and Vision (2013), Data reuse and the open data citation advantage. PeerJ 1:e175; DOI 10.7717/peerj.175

Our Institute provides a technology Solution to Data Sharing

An open-source software to share, cite, and find data.

Developed at Harvard's Institute for Quantitative Social Science

2006 (we started)

2017

dataverse.org

HOW Researchers SHare & Use data with dataverse

Harvard Dataverse Repository

> 70,000 datasets total
> 49,000 datasets uploaded to Harvard Dataverse repository

200 datasets/month

> 340,000 files

4,000 files/month

> 2.5 M downloads

60,000 downloads/month

Datasets Added

Downloads

dataverse.harvard.edu

King, 1995, Replication, Replication

Altman and King, 2007, A Proposed Standard for the Scholarly Citation of Quantitative Data

Altman et al, 2001, A Digital Library for the Dissemination and Replication of Quantitative Social Science

King, 2007, An Introduction to the Dataverse Network as an Infrastructure for Data Sharing

Crosas, Honaker, King, Sweeney, 2015, Automating Open Science for Big Data

Crosas, 2012, The Dataverse Network: an open source application for sharing, discovering, and preserving research data

Altman and Crosas, 2013, The Evolution to Data Citation: from principles to implementation

Crosas, 2013, A Data Sharing Story

2014, Joint Declaration of Data Citation Principles

Pepe et al, 2014, How Do Astronomers Share Data?

Goodman et al, 2014, Ten Simple Rules for the Care and Feeding of Scientific Data

Castro et al, 2015, Achieving Human and Machine Accessibility of Cited Data

Sweeney, Crosas, Bar-Sinai, 2015, Sharing Sensitive Data with Confidence: The DataTags System

Meyer et al. 2016, Data Publication with the Structural Biology Data Grid Supports Live Analysis

Wilkinson et al, 2016, The FAIR Guiding Principles for Scientific Data Management and Stewardship

Bierer, Crosas, Pierce, 2017, Data Authorship as an Incentive to Data Sharing

Our Contributions to Enhance data sharing

2017

Findable
Accessible
Interpoperable
Reusable

Data should be ...

Wilkinson et al. , 2016, "The FAIR Guiding Principles for Scientific Data Management and Stewardship" Nature Scientific Data

FAIR DATA in Dataverse

Data Files

Metadata

Data Licenses, User Agreements

Dataset Versions

Data Citation with Persistent Identifier (DOI)

A Dataverse is a container of Datasets and a Dataset is a container of data files, documentation, and code

Dataverse supports Astronomy Data

Supports default astronomy metadata fields (based on virtual observatory schema)
Extracts header metadata from FITS files upon ingest

Dataverse used by Max-Planck Institute ...

Dataverse in the astronomy news ...

More than 12,000 downloads!

What are we working on NOW?

Data Provenance

track the original source of a Dataset

Pasquier, Lau, Trisovic, Boose, Coutierer, Crosas, Ellison, GIbson, Jones, Seltzer, 2017, If These Data Could Talk, Nature Scientific Data (Data Provenance examples from CERN and Harvard Forest)

ClouD Dataverse

Combine data repositories with Cloud computing

Data Privacy

classify and handle datasets based on Their privacy level

Harvard Data Privacy Tools Project: privacytools.seas.harvard.edu

DataTags Project: datatags.org