Presented: 24 June 2020
Presenter: Matthias Liffers
#8 in the 8 webinar series of the FAIR data 101 training webinars.
Good afternoon, everybody. I’d like to welcome you to this last webinar for the FAIR data 101 course. First of all, I would like to acknowledge the traditional owners of the lands on which we all are. For me in Perth, that is the Nyoongar Whadjuk people, and I’d like to pay my respects to the elders past and present.
I’d also like to extend that respect to any members of first nations attending this webinar. All right. Yes, so the final webinar and we’ll be finishing up the reusable side of things. And then talking a bit about FAIR beyond data. I did say at the very beginning of the whole series that the FAIR guiding principles were not intended to be just about data, but I have to admit we’ve been spending a lot of time talking about data.
So we’ll be talking a little bit about other things that can be FAIRified as it were. I’d like to remind you all that the FAIR data 101 course is governed by a code of conduct. If you do observe any breach of the code of conduct, could you please report that to the ARDC via the form that is linked to from the code.
I’d also like to remind you all that you can enter your questions at any time in the question modules here in Go To webinar, and there will be a facilitated Q&A session after I’ve finished talking, basically. Liz will be joining us to do that. All right. As I said, we’re going to be finishing up the reusable side of things and I will mostly be talking about provenance.
Provenance of research data, not provenance of artworks or antiquities if you watch too much antiques roadshow like me. Then I’ll be talking about FAIR beyond data, software training materials. Then I’ll also be uploading a few options that are open to you if you want to continue your education or practice in FAIR. All right, let’s get right into it.
So at one point, too, as Liz described on Monday, metadata and data are associated with detailed provenance. All right, so that is a little bit problematic. In fact, getting detailed provenance can be tricky or rather it can be tricky but also it can be easy. Areas in which it is particularly tricky is when people work a lot with spreadsheets.
Specifically in a spreadsheet program like Microsoft Excel or Google sheets. Look, I’m absolutely guilty of this kind of behavior myself as well. I will get a data set and I will fire it up, will open it up in a spreadsheet program. I’ll mess around with it and I’ll move things around, clean things up, add columns, remove columns, stuff like that, and then save it and move on.
The issue there is I haven’t made a concerted effort to record what I’ve actually done to that spreadsheet. Unfortunately, Excel doesn’t really have a formal mechanism for recording that kind of thing either. In fact, the closest thing you have to that kind of recording mechanism is the undo redo buttons in Excel. And you tend to lose them … sorry, not just Excel.
I’m bashing Excel, I’m sorry. Any spreadsheeting software, the issue is that that undo, redo history can vanish when you save the file or when you’ve opened the file again from scratch. Now, there are other tools that you can use to work with tabular data that do save or permit you to save a nice detailed history. One of my favorite tools, for example, OpenRefine.
It allows you to open files, tabula data generally, manipulate that data and then you can, while saving that data again, you can also save the complete history of what you’ve done to that data. And in fact, you can load that history and apply it to different data sets, which is quite interesting.
If you’d like to learn more about OpenRefine, I strongly suggest that you have a look for the next available library carpentry course in your area, or in fact, online at the moment. Another way to record a detailed provenance of data manipulation is to separate the process from the data. Because really when we’re talking about recording a provenance, what we want is we want to know what the original data was. We want to know what the process was.
And we also want to know what the resulting data was and then link that all nicely through some kind of relationship. So you can explore, you can go backwards and forwards. I have a data set. How did that data set get to be in its current state? Look, there’s the process that was used to transform older data into this new state. And you can trace that through.
And then by investigating that process, you know that the data can be trustworthy. I’d like to remind you all of one of the activities that we asked you to do a few weeks ago, and that was the GLAM workbench tool or using a Jupyter notebook to interrogate the Trove API and get data from that and manipulate that data. Okay. So guess what? That was provenance.
So the entire Jupyter notebook programmatically defines the transformations, or indeed not just transformations, but first what data do we retrieve from the Trove API and how do we retrieve that through to any kinds of transformations to that data, and then displaying that data. And that is the full provenance. If you completed this activity, you essentially did record the provenance of the activity that you were doing. Fantastic.
So this is why I personally think … and look, it’s not just me personally, a lot of people would really like to encourage people to move away from using only Excel or sorry, spreadsheet program, to manipulate data in ways that aren’t easily then recorded or logged. And instead move to using some formal programming language. So in this Jupyter notebook here we used the Python language.
Python is one of the most popular languages for scientific computing these days. Another quite popular language is R, and hopefully you’ve heard of one or both of these. And by writing your transformations, by defining them programmatically in a script or a program and using that to analyze your data, you can not only keep that original data and not accidentally save over it like I’ve definitely done in the past using spreadsheets.
So you have that original data, you then have the process that you used to transform that data to script and that script outputs another file, the resulting data. And between the three of them, you have a nice provenance trail there. Another reason why it’s really nice to use software or to write software for data transformation is that software can then itself the made FAIR or available to others for potential reuse.
There are hundreds, if not thousands of scientific or research computing packages … Sorry, Andy, could you please mute yourself? I can hear you. So there are hundreds, if not thousands of research computing packages available in a number of different languages, Python, R, or even really old ones like FORTRAN or Cobalt.
And using these packages can help researchers save a lot of time because they don’t need to necessarily implement completely new packages from scratch themselves. They can look through a library of available packages and see if any are available to do the kind of data analysis that they need.
Now, one reason why we’d like to make these packages themselves FAIR … And in fact, with a lot of software, the focus is more on making it unlike FAIR, which doesn’t necessarily think too much about openness or rather it encourages openness, but doesn’t require it. For good provenance, that software probably really should be made open as well.
Because if the software isn’t open, then it’s a black box and you cannot inspect the internal workings of that black box to work out exactly what has happened with the data transformation. In fact, a colleague of mine was telling me how one very popular scientific computing program, SPSS, you can ask it to run any kinds of analysis on your data and for one particular analytical method called an ANOVA … now, don’t ask me what that is.
I’m sure I learned it in undergrad mathematics about 20 years ago, but I can’t tell you anymore. But there are several different ways of performing an ANOVA and unfortunately, SPSS doesn’t tell you which method it uses. So you tell it, “Please run an ANOVA.” And it will give you the results and say, “Here’s your ANOVA.” But it doesn’t give you the provenance of which analytical method was used.
Why this is important, or why we would like to be able to inspect research software and see how it works is that despite all best intentions, software can have bugs in it. So for example this python software package using the Willoughby Hoye method, it has been used in hundreds, if not thousands of computational chemistry projects and papers.
This particular script was found to rely on a computer operating system to do something. But unfortunately, the programmers didn’t realize that different computer operating systems when asked to perform the same function would do it differently. So under windows, under Linux, under older versions of Mac OS and new versions of Mac OS, what you would get from the operating system was subtly different.
It was to do with how files are ordered or lists of files are ordered. Different operating systems would order file lists in different ways. Now, the result of this particular bug was that under different operating systems, the same analysis on the same data would come up with different results, but subtly different results.
The results weren’t big enough to be noticeable to casual inspection by a human, but the results were different enough to make a big enough difference, a statistically significant difference to the results. And so this bug was only found because the software was opened and it could be investigated and scrutinized by other researchers.
We found the error, they reported that in a paper and the package has since been fixed and made stronger due to its open nature. All right, so that is one degree of providing a provenance trail, having the research software available for inspection. Now, the next level of recording provenance.
Now, there’s always a metadata schema and in this particular scale … sorry, and for provenance, there is a W3C recommendation and W3C recommendations, to be honest, are really standards. So the worldwide web consortium has this recommendation called PROV, or rather it’s a group of recommendations.
Now, I definitely will not go into too much detail about PROV because it’s pretty detailed and beyond the scope of an introductory course like this. But the basic idea of PROV is that things are described as linked data. Now, we’re all familiar with linked data, we went through that before. And that everything is one of, at the most fundamental level, three different kinds of objects.
So you have an entity and in this case, entity is data, a data set or a discrete data set, or data object. You also have activities and activities are processes by which data is transformed or generated. And then you have agents. And in the same way that the entity is the what, the activity is the how, the agent is the who, which person, which human, or which organization undertook the activity that produced an entity from another entity.
This can all be recorded as linked data in a variety of different formats. You can record it as XML or JSON or another language called Turtle. And there is an excellent primer on the W3C website on PROV and I strongly encourage you to have a read of that primer. And I think the first paragraph of that primer, which I found quite entertaining, is that you can engage with PROV as much or as little as you like.
That is to say, you can implement just a small part of PROV for your purposes, or you can try entering from the fire hose and implement all of PROV for your system. And PROV is being used in some organizations in Australia for recording provenance of their research data. But it is by no means universal.
I think recording provenance in a really detailed and systematic way is something that’s still to come and that we could all work on together as an Australian community, or even as a part of a worldwide community. All right, so that’s enough about provenance. Let’s get talking about something that isn’t data, although to be honest, I’ve just spent a lot of time talking about software, didn’t I?
Well, good news is I’m going to talk a bit more about software. So FAIR software. Now, when the FAIR guiding principles were written, it was imagined that they could apply to any kind of digital research output and to some degree, perhaps analog research outputs as well. Although that’s a little bit trickier.
Now, those guiding principles were originally written quite some years ago, and since then some very intelligent people have been thinking quite hard about what are the FAIR principles can be applied to software, but what of the FAIR principles don’t quite work or don’t fit together very well?
So there was a paper released recently by Lamprecht et al and they do acknowledge that software data are very similar, more similar than they are different in the way that software is a special kind of data. But there are some very significant differences between data and software as digital research objects that does require us as data stewards or digital research object stewards to treat them differently.
I’ll go over a couple of those things just very quickly. But this paper is a great read and just like the W3C PROV primer, I recommend that you have a bit of a read of this. And there is at least one Australian author on that paper as well. First up, licensing. Now, Liz discussed the creative commons license on Monday.
And so the creative commons license, we’re now up to version 4.0, but in the very early days of creative commons version 1.0, it wasn’t regionally for creative works, artworks, films, photos. Are photos artworks? Possibly. So things that are more artistic in nature rather than research based or scientific. In fact, the original creative commons license wasn’t even terribly good for research data.
And it took several iterations of creative commons before it started being relevant or before its particular clauses were made relevant for data as well. And so now we’re at creative commons 4.0, which is a good, robust license to be able to apply to data. Now, unfortunately, creative commons isn’t very applicable to software because of those differences between software and data.
At the same time, the Free and Open Source Software Movement has existed for a lot longer than the creative commons license has. And in fact, that community, the FOSS community has come up with many, many different kinds of software specific licenses. It’s probably in the same way that we like to joke, “There’s so many standards, we need to create a new standard that unites all the standards.”
And then all of a sudden we have yet another standard that needs to be united with everything else. So they are all these existing licenses and you might be familiar with things such as the Apache license, the MIT license, or the GNU, General Purpose License, GPL. These licenses are written from the get go or were written from the get go as software licenses and particularly open source software licenses.
So we strongly encourage you to investigate those and have a look at those rather than trying to apply a creative commons license to research software. Now, another way in which software and data differ quite a bit is that it is quite normal for software to be versioned and receive regular updates or continuous improvement, to use a management buzzword.
These updates happen quite regularly, whereas with data it’s not generally expected that a data set would have new versions. Some data sets do have new versions come out on a regular basis. So if you’re running a longitudinal study and you’re running a survey on a cohort every year, then there’s the new version of the data set as more data gets added.
However, not all research projects work like that. With software though, it absolutely is the case that new versions come out, there’s bugs that needed to be fixed, there is new functionality that needs to be added, or the software might start using a different method of calculating things. And so in order to link all the different versions of software with each other in the metadata records there are versioning practices that can be applied to keep things together and to aid citation of that software.
All right, now I’m definitely done talking about software and I will now start talking about FAIR trading material. This might be something that’s little closer to your hearts as I know that many people attending this course are themselves trainers and might in fact be seeking to train others in FAIR data practices. And quite recently, there was a paper, I think it was written by once again, a group of very clever people.
And they came up with 10 simple rules for making training materials FAIR. This illustration, this diagram I find fantastic because they created it for this particular paper. Realize now I’m citing the image here and not the paper. So I will provide the DOI of the paper itself once I’m done with this webinar. It’s another great read and probably directly relevant to a lot of people in this course.
So going clockwise around this diagram … well, they haven’t put all of the 10 simple rules on this diagram. But you can see how there are some things, some actions you can take with the training materials you create and those actions can assist more than one of the aspects of FAIR. So there’s a fair bit of overlap between find-able and accessible. Although then … yeah, sorry, interoperable and reusable stand by themselves. Oh wait, there’s number one, right in the very middle, share.
I think this is a really good practical guide for those of us who do develop into a little bit of training to help us make our training materials as useful to others as possible. Not just to those who are learning using that training material, but also people who might like to train using that training material. And then I have a bit of a navel gazing question for you, is FAIR data 101 fair?
Now, this is one of the topics for our community discussions next week and you’re absolutely welcome to completely sledge us basically … well, sorry. Be mindful of the code of conduct, challenge ideas, not people. So if you can give us some feedback on whether you think FAIR data 101 is fair, and if it’s not, what could we do differently?
All right, so that’s it. Now, that is the end of the course material or rather the end of the topic, the FAIR related webinars. You might be wondering now, keeping that it looks like certainly for some parts of Australia lockdowns might continue for a little bit longer, so we might have a bit of spare time, where to from here?
And so we’ve come up with some suggestions of avenues you might like to pursue depending on your particular interests. First up, Liz and I spent a lot of time talking about community agreed standards. And the first suggestion we have is that you join one of the communities that agree on the standards.
Now, the two most disciplined agnostic communities were around the world, there’s the Research Data Alliance and there’s Force11. The Research Data Alliance, it’s a worldwide body and the ARDC contributes to the Research Data Alliance. We assist in its governance. And within the Research Data Alliance, there were, I checked this morning, there are approximately 200 groups and each of these groups … so there are two kinds of groups.
There are interest groups, and that’s a long term group. People get together and discuss topics of common interest. And then there are working groups. And the idea for a working group is that it is mobilized for a year or two, and it’s formed to produce a particular output. So whether that is a framework or a series of recommendations, so that group creates this thing, publishes it, releases it to the world for free and disbands.
And the Research Data Alliance has, I never get this right, biannual, every twice a year it has a plenary somewhere in the world. And we were meant to have one in Australia earlier this year, but hey, guess what? That didn’t end up happening, but it did happen virtually.
At these plenaries every six months, the recommendations of working groups that are winding up are released and people can get together and discuss what the next piece of work could be, what new working groups could be formed. Force11 is similar but thinking more about research communications and scholarship. And they also collaborate on many things.
So that’s to join the communities that agree on the standards. An example of a RDA and joint Force11 and RDA working group is this FAIR 4 Research Software Working Group. Sorry, I’m not done talking about software. So this group is still in its forming stage and if you have a look at the chairs, we actually have two Australians, Michelle Barker and Paula Martinez who chairs.
And the other chairs come from around the world. So Dan Katz is from the U.S. Neil Chue Hong is in the UK. And the idea for this working group once it is endorsed, it’s not yet endorsed but hopefully soon, is to come up with concrete FAIR principles that are really tailored towards software. In fact, I have a couple of webinars next week, I believe, I’ll be able to share the details with you.
One is at a very Australia friendly time. So there are many, many working groups. This is just one of them. So there might be a working group that covers your particular domain interest. So sorry, I did say that RDA is domain agnostic. But there are many domain specific groups within the RDA.
So research out of agriculture, for example. Okay. And something else you might be interested in doing is working on or developing some real practical FAIR skills. Now, the really big one coming up soon is the banner up across the top, FSCI, the Force11 Scholarly Communication Institute.
This normally happens in the U.S. so it is a bit inaccessible for us here in Australia, but due to worldwide circumstances, it is happening online this year, which means you don’t need to travel to the U.S. but you still need to be awake at the appropriate time. And given the U.S. is approximately 12 hours’ time difference, that could be a bit of a challenge to attend.
But at least it’s possible, right? A little closer to home. Coming up soon on … Yeah. Sorry. FSCI is in the first two weeks of August. A bit close to home down the bottom there, Hacking Heritage, the GLAM workbench. So if you enjoyed that exercise, accessing the Trove API with a Jupyter notebook, Tim Sherratt at Canberra University or University of Canberra, he is running a three day course delivered online, of course.
You can sign up for that and that’s happening in the coming weeks as well. In the coming month at any rate. There is also a Carpentries Workshops which you may or may not have heard of before. The Carpentries is both an organization and a suite of lessons designed to train researchers and research support professionals in research computing methods in a very accessible and very friendly way.
The focus is on making sure that everyone has a fulfilling learning experience and everyone grows together. In fact, we got the idea for our code of conduct from The Carpentries. Then there were also some online resources available. So Foster is a European group about open science, and they have a number of resources, open science handbook, for example.
And similarly, The Turing Way from the alan-turing-institute also has a handbook rather, The Turing Way is the open science handbook and full of interesting lessons and theories and ways to make research more FAIR, although they do probably focus more on the concept of open science rather than FAIR research. Okay. So that’s it for things that you can do for yourself in terms of continuing your learning.
For this particular course, the wrap up now is that we have activities, community discussions and the quiz. So the activities and the quiz will be released sometime tomorrow. We’re just finalizing those, making sure that we haven’t gotten any spelling errors. The community discussions continue as normal next week, and the topics are already available.
Then stickers, we will be sending out stickers. The problem is we don’t know where you are. So we need to tell us where to post your stickers to, and in order to be able to tell us, you’ll need to complete a feedback form. The feedback form itself is anonymous, but we have separated but we’ve separated them, we’ve got a separate survey asking you where you live, so we can’t link those results together.
So you can be as brutal as you want in the feedback, but remember, please be kind as well. And then let us know where you would like your sticker sent to for completing the course. We also have some bonus content available for you. So the ARDC is partnering with CAUL, The Council of Australia University Librarians, and the AOASG, the Australasian Open Access Strategy Group, to deliver a webinar on FAIR beyond data.
Now, they’ll probably be focusing more on research publications. They’re my favorite software. But I heartily recommend to you to sign up. That’ll be next week on Monday at a very Australia friendly time, West Australia and East Australia. Yes, I think that’s all correct. Go to ardc.edu.au/events to register there. Now, that is it for me. So Liz, are you there?
Yes, I am here.
Hurray. Okay. How many questions do we have?
We have a few comments, people piping up about ANOVA. But the first question, which I have a half typed out bunch of links to share, but we’ll just ask the question. Can you recommend any good resources for choosing the right license for software, Matthias? But you can throw that back to me.
Yes. There is a website and I believe it is called something as simple as choose a license. Yes, choosealicense.com. Now, this website was created by GitHub and GitHub is owned by Microsoft. However, Microsoft has really gotten on board the open source train recently, which is quite nice.
And this is a very friendly way of picking out an appropriate license for your software. Bearing in mind that this was created in the U.S. Now, the ARDC is working on a software licensing guide. I can’t give you an ETA, unfortunately. But we would like that to be available sooner rather than later. All right.
Next one. Open to me.
No, there are no other questions at the moment. So I do encourage people to ask … look, maybe we can extend this Q&A out beyond just focused solely on reusability, if you would like to ask questions from this last cumulative series of weeks, and happy to wait while you form those questions in your mind and compel them to us via the question or chat module.
Yes, I see. I know the analysis of variance. That’s the one that was … Yeah. It’s honestly going back 20 years for me since I did first year statistics in university, and I can’t believe I’m that old already. All right, waiting patiently, type them out. There will, of course be opportunity for further discussion in the community discussions next week or … sorry, my earphones keep trying to fall out. Or there is the Slack, if you want to ask questions there, we’ve had some great discussion recently.
So discussing the data publishing and embargoes, there was an article about COVID-19 or the perils of people ignoring metadata standards when sharing COVID-19. That was quite interesting, wasn’t it, Liz? You were telling me how … what was it? You said the lack of appropriate metadata to do with the provenance, in fact, not saying where or when or how that data was collected made that data completely useless.
Yes, that’s right. In the circumstances of a pandemic, you really want to make sure that the metadata you have for the location and the time at which disease had been reported is complete. But these researchers have found lots of instances where that information has not been complete, which makes that research data absolutely useless until somebody has completely gone through and filled it out. So nothing like a pandemic to sharpen the inequalities of data quality.
All right. So we just killed some time there. Have any more questions come in at all?
Oh, I guess. Hang on. There’s a request to email the slide shows two course participants, which we will undertake. We’ll put the slides out on the Slack channel. But we can see if we can send an automatic email to everyone. And there’s one final question about how best to find out about Carpentries events.
Okay, great. Now, the problem with Carpentries events, so the carpentries does really focus very much on face-to-face training, which is a problem right now. So they’ve been working very hard to pivot to … sorry, using another management buzzword word there, to deliver more online training. And so there has been a bit of a bump in availability of training workshops.
But there are several places you can check. You can go to The Carpentries website to see if there are any workshops coming up soon that you can attend. In fact, with more of them being online, it might be easier to find one you can attend, although it might not be at a very friendly time of day. Alternatively, there could be already somebody at your institution delivering Carpentries workshops.
I know some universities like Macquarie University, Monash University, and that’s just the tip of the iceberg, they are delivering regular Carpentries workshops for their staff and students. And then there are also some organizations in Australia who will deliver Carpentries workshops for you. Intersect, they’re based in Sydney, but they are active around the country.
They can, for a fee, deliver a Carpentries workshop for you. Alternatively, I am one of the Carpentries regional coordinators for Australia, and I welcome you to get in touch with me to ask about your specific institution or area if you’re not sure where to start.
Excellent. Few more questions now. And also another helpful reminder that QCIF also delivers Carpentries workshops online. And so thank you for submitting the question. I’m just going to cut and paste that into the chat as well.
Yes, so QCIF tends to deliver only to their members and their member organizations are mostly universities in Queensland, whereas Intersect will also deliver some non-member organizations.
I do have a final question. Is there guidance available from ARDC on adding DOIs to research outputs in repositories?
Absolutely. No, I’m sure there is. We’ll share a link with you. We help facilitate getting DOI or helping Australian institutions get to DOIs for their research outputs. Now we are limited to research data and so non publication research objects generally. We also have scope to help facilitate DOIs for great literature.
But when it comes to formal publications, like say if your institution publishes a journal and you want to get DOIs for those journals and those journal articles, you’re probably best off going directly to Crossref, which is the organization that produces most of those DOIs.
Yeah. That’s true. I think that in answer to that question, you’d be looking at contacting the data and services team at ARDC. Let me just put a link to the part of our website that refers to our DOI service. And follow up, I’ll put the team to contact, the person to contact into the Slack channel if that’s okay.
There is also your local ARDC liaison. So we have at least one engagement’s officer in each state and you’re welcome to contact them and ask them if you know them, if you don’t know them, ask us.
Yep. So Matthias, one final question and then I’m going to put a hard line underneath all of these because I need to run off and pick up a certain child from school. Do you have any advice or useful links for someone who would like to learn Python or R who has never done it before?
Yes. So The Carpentries is a fantastic way to … if there’s a workshop available to you, attending a software Carpentries course on Python or R is a fantastic and friendly introduction. It’s gentle and it really works with real world examples. It’s not a high level theoretical overview of programming as you would get in a computer science or software engineering course, but it’s very practical and gentle introduction to coding.
Yes. And I would add that, so it’s checking out with your organization if they do have any Carpentries instructors who are running workshops virtually, but also you might consider things if there might be a Rezbaz activity in the future, which tend to be festivals.
But mostly have been canceled, I believe this year, certainly in Australia. But these are good touch points to get in touch with and learn Python and R normally through the Carpentries in your state. So it’s likely that there might be some kind of virtual activity this year or some kind of face-to-face thing next year.
Yeah. There were also a number of MOOCs available. So data science being the hot new thing. Well, it’s not that new anymore, but data science being the hot new thing. That’s the career of the future at the moment. There are quite a few open and free to attend MOOCs. So one I did was from Johns Hopkins University in the U.S.
All right. So I think that we have run out of time, Liz.
Yeah, would you like to wrap up?
Well, I don’t actually have anything more to say, but I think it’s nice if we both stay on and we both give a wave, say thank you to everybody for attending these webinars. We look forward to seeing you in the community discussions next week and also on Slack as well. And we will be sharing a lot of information with you after this presentation. There’s lots of links.
Sorry, something I completely forgot to mention was the ARDC’s communities of practice. So the ARDC runs or is affiliated with a number of communities of practice on different topics. And we strongly encourage you to check out the ARDC communities of practice page to see if there is a community there that suits you. For example, many people in this course might be interested in the data librarians’ community of practice if you’re not already a member.
All right. Now, I think that’s enough. So definitely thank you very much for joining us for the past seven weeks or it’ll be eight weeks next week. It has been an amazing time and it’s been absolutely fantastic being able to deliver these webinars to you. If you do have any questions, please let us know. Otherwise, we will see you next week. Thank you.