Bank Loan Approval Analysis: A Comprehensive Data Analysis Project
Learn to speak open
1. Learn to speak Open: What, Why, How
in Open Science
Iryna Kuchma, EIFL
Sarah Jones, Digital Curation Centre, University of Glasgow
OpenCon Nairobi 2016
2. WHAT DOES IT MEAN
TO BE OPEN?
Image by biblioteekje CC-BY-NC-SA www.flickr.com/photos/biblioteekje/3992172265
3. Open Science implementation – is it a wrap rage?
Image from Wikipedia https://en.wikipedia.org/wiki/Wrap_rage
4. What is open science?
“science carried out and communicated in a
manner which allows others to contribute,
collaborate and add to the research effort, with
all kinds of data, results and protocols made
freely available at different stages of the
research process.”
Research Information Network, Open Science case studies
www.rin.ac.uk/our-work/data-management-and-curation/
open-science-case-studies
5. Open methods
• Documenting and sharing workflows and methods
• Sharing code and tools to allow others to reproduce work
• Using web based tools to facilitate collaboration and
interaction from the outside world
• Open notebook science – “when there is a URL to a
laboratory notebook that is freely available and indexed
on common search engines.”
http://drexel-coas-elearning.blogspot.co.uk/2006/09/open-notebook-science.html
6. Open access to publications
• Free, immediate, online access to the results of research,
coupled with the right to use those results in new and
innovative ways
• Make sure anyone can access your papers
Strategies to achieve open access:
open access journals and monographs
open access repositories
Find out what your publisher allows on SHERPA RoMEO
– www.sherpa.ac.uk/romeo
7. Open data
make your stuff available on the Web (whatever format) under an open licence
make it available as structured data (e.g. Excel instead of a scan of a table)
use non-proprietary formats (e.g. CSV instead of Excel)
use URIs to denote things, so that people can point at your stuff
link your data to other data to provide context
Tim Berners-Lee’s proposal for five star open data - http://5stardata.info
“Open data and content can be freely used,
modified and shared by anyone for any purpose”
http://opendefinition.org
8. The research workflow – a cyclic model
preparation
analysis
writingpublication
outreach
assessment discovery
10. https://innoscholcomm.silk.co
101 Innovations in Scholarly Communication
How can libraries support changing research workflows?
Bianca Kramer & Jeroen Bosman
EIFL General Assembly, Riga, November 13, 2015
@MsPhelps
@jeroenbosman
15. Science as an open enterprise
https://royalsociety.org/policy/projects/science-public-enterprise/Report
“Much of the remarkable growth of
scientific understanding in recent
centuries is due to open practices; open
communication and deliberation sit at
the heart of scientific practice.”
Royal Society report calls for ‘intelligent
openness’ whereby data are accessible,
intelligible, assessable and usable.
16. Some benefits of openness
• You can access relevant literature – not behind pay walls
• Ensures research is transparent and reproducible
• Increased visibility, usage and impact of your work
• New collaborations and research partnerships
• Ensure long-term access to your outputs
• Help increase the efficiency of research
17. Saving wasted time
OA helps to reduce time spent finding/accessing material:
“If around 60 minutes were characteristic for researchers
(the average time spent trying to access the last research
article they had difficulty accessing), then in the current
environment the time spent dealing with research article
access difficulties might be costing around DKK 540
million (EUR 72 million) per year among specialist
researchers in Denmark alone.”
Access to research and technical information in Denmark,
Houghton, Swan & Brown (2011)
http://eprints.ecs.soton.ac.uk/22603
18. Cut down on academic fraud
www.nature.com/news/2011/111101/full/479015a.html
19. Validation of results
www.guardian.co.uk/politics/2013/apr/18/uncovered-error-george-osborne-austerity
“It was a mistake in a spreadsheet that could
have been easily overlooked: a few rows left
out of an equation to average the values in a
column.
The spreadsheet was used to draw the
conclusion of an influential 2010 economics
paper: that public debt of more than 90% of
GDP slows down growth. This conclusion was
later cited by the International Monetary Fund
and the UK Treasury to justify programmes
of austerity that have arguably led to riots,
poverty and lost jobs.”
20. Acceleration of the research process
“As more papers are deposited and more scientists
use the repository, the time between an article being
deposited and being cited has been shrinking
dramatically, year upon year. This is important for
research uptake and progress, because it means that
in this area of research, where articles are made
available at – or frequently before – publication, the
research cycle is accelerating.”
Open Access: Why should we have it? Alma Swan
www.keyperspectives.co.uk
22. A citation advantage
A study that analysed the citation counts of 10,555 papers on gene
expression studies that created microarray data, showed:
“studies that made data available in a public repository
received 9% more citations than similar studies for
which the data was not made available”
Data reuse and the open data citation advantage,
Piwowar, H. & Vision, T. https://peerj.com/articles/175
23. Increased use and economic benefit
Up to 2008
• Sold through the US Geological
Survey for US$600 per scene
• Sales of 19,000 scenes per year
• Annual revenue of $11.4 million
Since 2009
• Freely available over the internet
• Google Earth now uses the images
• Transmission of 2,100,000
scenes per year.
• Estimated to have created value for
the environmental management
industry of $935 million, with direct
benefit of more than $100 million
per year to the US economy
• Has stimulated the development of
applications from a large number of
companies worldwide
The case of NASA Landsat satellite imagery of the Earth’s surface:
http://earthobservatory.nasa.gov/IOTD/view.php?id=83394&src=ve
24. So what needs to change?
Conclusions from Emilio Bruna:
• Develop a better system of incentives from the
community for archiving data and code
• Teach our students how to do this NOW - it’s much easier
if you develop good habits early
We need to stop telling people “You should” and get
better at telling people “Here’s how”
25. HOW TO PRACTICE OPEN SCIENCE
Image by Paul Downey CC-BY www.flickr.com/photos/psd/3925801816
29. Make a list of open access publication
options in your particular
field. Chances are you will be
surprised by the range of possibilities
Erin McKiernan
30. Discuss access issues with your
collaborators up front, before the
research is done and the articles
written.
Erin McKiernan
31.
32.
33.
34.
35. Blog about your science, and in language
that is comprehensible to non-
scientists. Doing this can ultimately
increase the impact of your work and can
even lead sometimes to press coverage
and to better press coverage.
Erin McKiernan
36. Be active on social media. This is the
way academic reputations are built
today, so ignoring the opportunities
presented is unwise.
Erin McKiernan
37. If for some reason you do publish a
closed-access article, remember that you
can self-archive a copy of your article in a
disciplinary or institutional or shared
repository.
Erin McKiernan
38.
39. Deposit in your local repository!
• Speak to the library
• Consider other relevant repositories for your field too
e.g. Arxiv - http://arxiv.org
• Check OpenDOAR for examples -
http://www.opendoar.org
40.
41. ~78% of publishers allow authors to openly archive a version of their published
manuscript: Breakdown of archiving policies from over 2,100 publishers. Source: Data
from SHERPA/RoMEO. Accessed October 2015 and plotted by E.C. McKiernan (CC BY)
46. 1. It’ll spoil my publication chances later
Well, it might, but in a recent survey only 7% of
institutions cited this as a frequent concern
amongst their students, and no concrete
examples were found of publication being
refused because the PhD thesis had been added
to an open access repository. If reassurance is
needed, then an embargo period can be applied,
with may be the record plus abstract still being
available to all.
47.
48. How to make data open?
1. Choose your dataset(s)
• What can you may open? You may need to revisit this step if you
encounter problems later.
2. Apply an open license
• Determine what IP exists. Apply a suitable licence e.g. CC-BY
3. Make the data available
• Provide the data in a suitable format. Use repositories.
4. Make it discoverable
• Post on the web, register in catalogues…
https://okfn.org
Initially I want to consider what it means to be open
The Research Information Network ran a series of open science case studies a few years ago to look at openness across different disciplines. These included astronomy, bioinformatics, chemistry, clinical neuroimaging, epidemiology and language technology.
Through this they came up with the definition. They defined open science as … The first aspect is about conducting science in an open way to encourage others to engage with it, and also sharing the outputs at various stages.
I’ll unpack this definition over the next few slides
Part of open science is working openly – having open methods. This involves documenting workflows and procedures so others can find and reuse your methods, sharing code and tools so others can reproduce your work, and using web based tools such as blogs and wikis to facilitate collaboration and interaction with others.
You may come across the terms ‘open notebook science’. This was coined by Jean-Claude Bradley. It’s about having an online labnotebook that others can find and follow.
Open science is also about providing open access to publications. This is well-established concept now, so one you’re probably familiar with.
Open access is about ensuring free, immediate, online access to the results of research. There are two main routes to ensuring open access – publishing in open access journals self-archiving a copy (usually a pre-print version) in a repository.
You can see what your publisher allows via Sherpa RoMEO – search for the journal title and it will tell you if there are gold, paid-for options, or what is permitted in terms of green routes (i.e. what version can be archived and under what embargo)
Open science is also about making data available. Open data is data that can be used, modified and shared by anyone for any purpose. So for example it can be mined, exploited, reproduced, commercialised... Data under a non-commercial license wouldn’t be open as you’re restricting how it can be used and by whom.
There are different levels of openness. Making content available on the web, in any format, under an open license is open data. It’s not necessarily that reusable though. By making it available as structured data, in non-proprietary formats, using URIs and linking to other data for context, makes it much more valuable.
Motiverend en voor ons belangrijk, afgelopen 20 jaar en vooral laatste 5-10 jaar meer in onderzoek/publiceren veranderd dan 350 jaar daarvoor
So we’re talking about being open at every stage, right from designing your research to conducting experiments and analysis, publishing results and releasing the associated outputs.
There are tools that can help at various stages and I’ve listed a few here. Taverna for example, help with documenting workflows, while MyExperiment is a VLE that allows you to share workflows and collaborate. LabTrove is a blogging platform that integrates with instruments so you can automatically publish documentation, and OpenWetWare is a wiki to share information, know-how, and research protocols in biology & biological engineering. GitHub is a code repository to handle versioning and allow others to collaborate / reuse code. RoMEO is a lit of journal policies, re3data.org is a list of data repositories and DataCite and Impact story help you to track the citation, reuse and impact of your work.
In the second half of the talk I’ve focus on reasons why you should be open – specifically the benefits and rewards
I like the quote from Ewan Birney that it was never acceptable to publish papers without making data available. If you can evidence and validate your findings, how do others know that they’re right?
The Royal Society report ‘Science as an Open Enterprise’ makes a similar point. This emphasises that much of the growth of scientific understanding is due to open practices. Being open about your work and encouraging feedback from others is at the heart of scientific practice.
The report calls for ‘intelligent openness’ – data shouldn’t just be accessible, they need to be intelligible by others so they can assess and reuse them.
What are the benefits to openness?
When publications are open, it means you can access all the relevant literature regardless of whether you have a subscription or not.
Open sharing also means that research is transparant and reproducible – you can validate results
By sharing you will increase your visibility and the reuse of your work, which in turn leads to more collaboration and partnerships
By depositing in repositories, you also safeguard your outputs since these are properly managed services that will curate your papers and data
Sharing in general also makes research more efficient – it’s quicker and new discoveries are made faster
There was a study done in Denmark on how open access saves time. They looked at the average time wasted looking for articles that were difficult to locate or gain access to (average = 60 mins per article), and extrapoloated this up to see what it was costing the sector annually. The estimate is €72 million in Denmark alone.
Sharing data also cuts down on academic fraud. You’ve probably come across the case of Diederik Stapel – a former professor of social psychology at Tilberg University in the Netherlands. He had literally been making up data to support his hypotheses and various headlines had been picked up by the media e.g. meat eaters are more selfish than vegetarians. Some students questioned the results and the whole story unravelled, showing this had been going on for years. What’s troubling is that this had gone unnoticed. The data weren’t made available and checked, so such a history of fraud could develop.
The validation of results is key. This article references an inadvertent error in a economics paper by Reinhadt and Rogoff. Missing some rows out of an average gave drastically different results – what was published suggested that countries with 90% debt ratios see their economies shrink by 0.1%. Instead, it should have found that they grow by 2.2% – less than those with lower debt ratios, but not a spiralling collapse. This mistake wasn’t picked up on initially as the data hadn’t been shared. The mistake fed into government policy - the findings of this paper were used as justification for austerity measures in the UK and various other countries in the EU.
To look at things in a more positive light, this study shows how open access publishing accelerates the research process. It looks at citations rates for articles deposited in arXiv – a preprints repository in maths, physics, astronomy, computer science etc The time lag between deposit and citation of an article has been shrinking drastically year on year. Nowadays the research cycle in High-Energy Physics (HEP) is approaching maximum efficiency as a result of the early and free availability of articles that scientists in the field can use and build upon rapidly.
Certain research communities have also seen the benefit of sharing data as it speeds up the process of discovery. This article shows how researchers in the field of Alzheimer’s research have agreed as a community to share data immediately to make scientific breakthroughs.
There’s also a citation advantage for individual researchers. This study by Heather Piwowar and Todd Vision looked at 10,555 paper of gene expression studies that had shared the associated microarray data. Those studies that shared data received 9% more citations.
There’s also an economic benefit, as seen by the case of the NASA landsat satellite images. These were sold until 2008 for $600 a scene. Now they’re freely available and used by Google Earth. Previously they sold 19,000 images a year, whereas now they transmit 2.1 million. The revenue has gone up incredibly too from $11.4 million to an estimated value of $935 million with direct benefit of more than $100 million. The release has also stimulated the development of applications from companies worldwide.
This case study comes from the Royal Society Report on Science as an Open Enterprise.
He suggests what needs to change and I think this is quite pertinent for today.
We need a better system of incentives so you get more tangible returns from openness – better chances on grant applications, more likelihood of promotion… This is likely to evolve during your careers.
If we teach people what to do now, it will become part of your processes and be easier in the long run. The learning curve won’t be as steep.
We also need to focus not on why you should be open, but how – and that’s what today’s presentations should give you.
I mentioned earlier about conducting science in the open. UsefulChem is an example of that. It’s a wiki platform where chemists can document their experiments. It’s effectively an open labnotebook. You can see here that the researcher has shared his aims, procedures, results, conclusion and a log of activities. There is also an option to comment.
My Experiment is like a Virtual Research Environment. You can share scientific workflows that others can comment on and reuse. You can see in the sidebar that it’s clear how the workflow is licensed and who credit should be given to if reusing. There are also a number of social media features too e.g. ratings, favourites and stats for views and downloads.
Speak to your library as there is usually an institutional repository you can use. There may also be relevant repositories given your field e.g. Arxiv for maths, computing science, physics etc.
This is the SHERPA RoMEO service. You can search for your journal title and then see what is allowed.
In this example, the author can archive a pre-print but not the publisher’s final version, and certain restrictions apply e.g. a 6 month embargo.
The final aspect to discuss is open data. The Open Knowledge Foundation suggests four simple steps. First off you need to decide what to share. Not all data can be made openly available due to commercial restrictions or sensitivities. Once you’ve decided what to share, determine what IP exists and apply a suitable licence. You should then make the data available in a suitable format so others can bulk download it. Remember that for it to be useful you want to share appropriate metadata and documentation too. Using repositories is useful to make sure your data are properly managed and preserved for the long-term. The final step is to make your data discoverable. Put it online, tell others about it and add details to various registries so it gets found.