Is it OK to cite preprints? Yes, yes it is.
Should you cite preprints in your papers and should journals allow this? This is a topic that gets debated periodically. The most recent round of Twitter debate started last week when Martin Hunt pointed out that the journal Nucleic Acids Research wouldn’t allow him to cite them. A couple of days later I suggested that journals that don’t allow citing preprints are putting their authors’ at risk by forcing them not to cite relevant work. Roughly forty games of Sleeping Queens later (my kid is really into Sleeping Queens) I reopened Twitter and found a roiling debate over whether citing preprints was appropriate at all.
The basic argument against citing preprints is that they aren’t peer reviewed. E.g.,
and that this could lead to the citation of bad work and the potential decay of science. E.g.,
There are three reasons I disagree with this argument:
- We already cite lots of non-peer reviewed things in ecology
- Lots of fields already do this and they are doing just fine.
- Responsibility for the citation lies with the citer
We already cite non-peer reviewed things in ecology
As Auriel Fournier, Stephen Heard, Michael Hoffman, TerryMcGlynn and ATMoody pointed out we already cite lots of things that aren’t peer reviewed including government agency reports, white papers, and other “grey literature”.
We also cite lots of other really important non-peer reviewed things like data and software. We been doing this for decades. Ecology hasn’t become polluted with pseudo science. It will all be OK.
Lots of other fields already do this
One of the things I find amusing/exhausting about biologists debating preprints is ignorance of their history and use in other fields. It’s a bit like debating the name of an actor for two hours when you could easily look it up on Google.
In this particular case (as Eric Pedersen pointed out) we know that citation of preprints isn’t going to cause problems for the field because it hasn’t caused issues in other fields and has almost invariably become standard practice in fields that use preprints. Unless you think Physics and Math are having real issues it’s difficult to argue that this is a meaningful problem. Just ask a physicist
You are responsible for your citations
Why hasn’t citing unreviewed work caused the wheels to fall off of science? Because citing appropriate work in the proper context is part of our job. There are good preprints and bad preprints, good reports and bad reports, good data and bad data, good software and bad software, and good papers and bad papers. As Belinda Phipson, Casey Green, Dave Harris and Sebastian Raschka point out it is up to us as the people citing research to make professional judgments about what is good science and should be cited. Casey’s take captures my thoughts on this exactly:
So yes, you should cite preprints and other unreviewed things that are important for your work. That’s called proper attribution. It has worked in ecology and other fields for decades. It will continue to work because we are scientists and evaluating the science we cite is part of our jobs. You can even cite this blog post if you want to.
Thanks to everyone both linked here and not for the spirited discussion. Sorry I wasn’t there, but Sleeping Queens is a pretty awesome game.
UPDATE: For those of you new to this discussion, it’s been going on for a long time even in biology. Here is Graham Coop’s excellent post from nearly 4 years ago.
UPDATE: Discussion of why it’s important to put preprint citations are in the reference list
Data Analyst position in ecology research group
The Weecology lab group run by Ethan White and Morgan Ernest at the University of Florida is seeking a Data Analyst to work collaboratively with faculty, graduate students, and postdocs to understand and model ecological systems. We’re looking for someone who enjoys tidying, managing, manipulating, visualizing, and analyzing data to help support scientific discovery.
The position will include:
- Organizing, analyzing, and visualizing large amounts of ecological data, including spatial and remotely sensed data. Modifying existing analytical approaches and data protocols as needed.
- Planning and executing the analysis of data related to newly forming questions from the group. Assisting in the statistical analysis of ecological data, as determined by the needs of the research group.
- Providing assistance and guidance to members of the research group on existing research projects. Working collaboratively with undergraduates, graduate students and postdocs in the group and from related projects.
- Learning new analytical tools and software as needed.
This is a staff position in the group and will be focused on data management and analysis. All members of this collaborative group are considered equal partners in the scientific process and this position will be actively involved in collaborations. Weecology believes in the importance of open science, so most work done as part of this position will involve writing open source code, use of open source software, and production and use of open data.
Weecology is a partnership between the White Lab, which studies ecology using quantitative and computational approaches and the Ernest Lab, which tends to be more field and community ecology oriented. The Weecology group supports and encourages members interested in a variety of career paths. Former weecologists are currently employed in the tech industry, with the National Ecological Observatory Network, as faculty at teaching-focused colleges, and as postdocs and faculty at research universities. We are also committed to supporting and training a diverse scientific workforce. Current and former group members encompass a variety of racial and ethnic backgrounds from the U.S. and other countries, members of the LGBTQ community, military veterans, people with chronic illnesses, and first-generation college students. More information about the Weecology group and respective labs is available on our website. You can also check us out on Twitter (@skmorgane, @ethanwhite, @weecology, GitHub, and our blog Jabberwocky Ecology.
The ideal candidate will have:
- Experience working with data in R or Python, some exposure to version control (preferably Git and GitHub), and potentially some background with database management systems (e.g., PostgreSQL, SQLite, MySQL) and spatial data.
- Research experience in ecology
- Interest in open approaches to science
- Experience collecting or working with ecological data
That said, don’t let the absence of any of these stop you from applying. If this sounds like a job you’d like to have please go ahead and put in an application.
We currently have funding for this position for 2.5 years. Minimum salary is $40,000/year (which goes a pretty long way in Gainesville), but there is significant flexibility in this number for highly qualified candidates. We are open to the possibility of someone working remotely. The position will remain open until filled, with initial review of applications beginning on May 5th. If you’re interested in applying you can do so through the official UF position page. If you have any questions or just want to let us know that you’re applying you can email Weecology’s project manager Glenda Yenni at email@example.com.
Data Retriever 2.0: We handle the data so you can focus on the analysis
We are very exited to announce a major new release of the Data Retriever, our software for making it quick and easy to get clean, ready to analyze, versions of publicly available data.
The Data Retriever, automates the downloading, cleaning, and installing of ecological and environmental data into your choice of databases and flat file formats. Instead of hours tracking down the data on the web, downloading it, trying to import it, running into issues (e.g, non-standard nulls, problematic column names, encoding issues), fixing one problem, and then encountering the next, all you need to do is run a single command from the command line:
$ retriever install csv iris $ retriever install sqlite breed-bird-survey -f bbs.sqlite
or from R:
>>> rdataretriever::install('postgres', 'wine-quality') >>> portal_data <- rdataretriever::fetch('portal')
The Data Retriever uses information in Frictionless Data datapackage.json files to automatically handle all of the complexities of “simple” data for you. For more complicated complicated datasets, with dozens of components or major data structure issues, the Retriever uses Python scripts as plugins to handle the major data cleaning work and then automatically handles the rest.
To find out more about the Data Retriever checkout the websites, the full documentation, and the GitHub repositories for both the Data Retriever and the R Data Retriever package.
Expanded focus and name change
For those of you familiar with the EcoData Retriever, this is the same software with a new name. Challenges with the data end of the analysis pipeline occur across disciplines and our tools work just as well for non-ecological data, so we’ve started adding non-ecological data and changed our name to reflect that. We’d love to hear from anyone interested in leading a push to add data from another discipline or just interested in adding a single favorite dataset.
As part of this we’ve changed the name of the R package from
The 2.0 release includes a number of major changes including:
- Python 3 support (a single code base runs on both Python 2 and 3)
- Adoption of the frictionless data datapackage.json standard (replacing our old YAML like metadata system), including a command line interface for creating and editing datapackage.json files
- Add json and xml as available output formats
- Major expansion of the documentation and hosting of the documentation at Read the Docs
- Remove the graphical user interface (to allow us to focus that development time on wrappers for other languages)
- Lots of work under the hood and major improvements in testing
- Broaden scope to include non-ecological data
We are also in the process of releasing version 1.0 of the R package. This version adds the new features in the Data Retriever and also includes major stability improvements, in particular in RStudio and on Windows.
We also have a brand new website.
Upgrading to the new version (UPDATED)
To ensure the smoothest upgrade to the new version we recommend:
retriever reset scriptsfrom the command line
- Uninstall the old version of the EcoData Retriever
- Install the new version
retriever updatefrom the command line
Henry Senyondo is the lead developer for the Data Retriever and has done an amazing job over the past year developing new features and shoring up the fundamentals for the software. He lead the work on 2.0 start to finish.
Akash Goel was a Google Summer of Code student with the project last summer and was responsible for the majority of the work adding Python 3 support and switching the project over to the
Dan McGlinn, the creator of the R package, has continued his excellent leadership of the development of this package. Shawn Taylor, a new contributor, was instrumental in solving the stability issues on Windows/RStudio.
In addition to these core folks our growing group of contributors to both projects have been invaluable for adding new functionality, fixing bugs, and testing new changes. We are super excited to have contributions from 30 different people and will keep working hard to make sure that everyone feels welcome and supported in contributing to the project.
The level of work done to get these releases out the door was only possible due to generous support of the Gordon and Betty Moore Foundation’s Data Driven Discovery Initiative. This support allowed my group to employ Henry as a full time software engineer to work on these and other projects. This kind of active support for the development and maintenance of research oriented software makes sustainable software development at universities possible.
New release of the EcoData Retriever
We are very exited to announce the newest release of the EcoData Retriever, our software for automating the downloading, cleaning, and installing of ecological and environmental data. Instead of hours or days trying to get complicated datasets like the Breeding Bird Survey ready for analysis, the Retriever lets you simply click a button or run a single command from R or the command line, and your computer does the rest.
It’s been over a year since the last retriever release and there are lots of new features and improvements to be excited about.
- We’ve added 21 new datasets, including major ecological and environmental datasets like eBird, Vertnet, and the Global Wood Density Database, and the PRISM climate data.
- To support all of these datasets we’ve added support for additional data types including greater than memory archive files, and we’ve also improved the ability to control where downloaded files are stored and how they are clustered together.
- We’ve significantly improved documentation and now have a new automatically built documentation site at Read The Docs.
- We’ve also made a lot of under the hood improvements.
This is also the first release that has been overseen by Weecology’s new software engineer, Henry Senyondo. We’re excited to have Henry on the team, and now that he’s around development of both the EcoData Retriever and other lab software projects will be happening more quickly.
A big thanks to the Gordon and Betty Moore Foundation’s Data-Driven Discovery Initiative for funding this development through Grant GBMF4563 and to the National Science Foundation for funding as part of a CAREER award to Ethan White.
UPDATE: Led by Dan McGlinn we also released a new version of the ecoretriever R interface for the Retriever last fall. This makes using the Retriever from R as simple as:
data <- ecoretriever::fetch("BBS")
Thoughts on preprints and citations
A couple of months ago Micah J. Marty and I had a twitter conversation and subsequent email exchange about how citations worked with preprints. I asked Micah if I could share our email discussion since I thought it would be useful to others and he kindly said yes. What follows are Michah’s questions followed by my responses.
Right now, I am finishing up a multi-chapter Master’s thesis and I plan to publish a few papers from my work. I may want to submit a preprint of one manuscript but before I propose this avenue to my advisor, I want to understand it fully myself. And I have remaining questions about the syntax of citing works when preprints come into play. What happens to a citation of a preprint after the manuscript is later published in a peer reviewed venue?
At the level of the journal nothing happens. So, if you cite a preprint in a published ms, and that preprint is later published as a paper, then the citation is still to the preprint. However, some of the services indexing citations recognize the relationship between the preprint and the paper and aggregate the citations. Specifically, Google Scholar treats the preprint and the published paper as the same for citation analysis purposes. See the citation record for our paper on Best practices for scientific computing which has been cited 49 times, but the vast majority of those are citations to the preprints.
Here’s an example with names we can play with: Manuscript 1 (M1) may require some extra analysis, but it presents some important unexpected results that I would like to get out on the table as soon as possible. M1 is submitted to PeerJ Preprints and accepted (i.e., published online as a preprint with a DOI). M2 is submitted to Marine Ecology Progress Series (MEPS) for peer review, and M2 cites the PeerJ Preprint M1.
Just a point related to vocabulary, I wouldn’t typically think of the preprint as being “accepted”. Any checking prior to posting is just a quick glance to make sure that it isn’t embarrassingly bad, so as long as it’s reasonably written and doesn’t have a title like “E is not equal to mc squared” it will be posted almost immediately (within 48 hours on most preprint servers).
1) Are preprints considered “grey literature”? That is, is it illegitimate for M2 to cite a work that has not been peer reviewed?
Yes, in the sense that they haven’t been formally peer reviewed prior to posting they are similar to “grey literature”. Whether or not they can be cited depends on the journal. Some journals are happy to allow citing of preprints. For example, this recent paper in TREE cites a preprint of ours on arXiv. Their paper was published before ours was accepted, so if it wasn’t for the preprint it couldn’t have been cited.
2) Is there a problem if M1 is eventually published in a peer reviewed journal but the published article of M2 cites only the PeerJ Preprint of M1?
I would say no for two reasons. First, assuming that M2 is published before M1 then the choice is between having a citation to something that people can read, science can benefit from, and that can potentially be indexed (giving you citation credit) vs. a citation to “Marty et al. unpublished data”, which basically does nothing. Second, all preprint servers provide a mechanism for linking to the final version, so if someone finds the preprint via a citation in M2 then that link will point them in the direction of the final version that they can then read/cite/etc.
In short, I think as long as you aren’t planning on submitting to a behind the times journal that doesn’t allow the submission of papers that have been posted as preprints (and the list of journals with this policy is shrinking rapidly) then there is no downside to posting preprints. In the best case scenario it can lead to more people reading your research and citing it. The worst case scenario is exactly the same as if you didn’t post a preprint.
EcoData Retriever now supports R and environmental data, and has more datasets
We are very excited to announce the newest release of our EcoData Retriever software and the first release of a supporting R package, ecoretriever. If you’re not familiar with the EcoData Retriever you can read more here.
The biggest improvement to the Retriever in this set of releases is the ability to run it directly from R. Dan McGlinn did a great job leading the development of this package and we got ton of fantastic help from the folks at rOpenSci (most notably Scott Chamberlain, Gavin Simpson, and Karthik Ram). Now, once you install the main EcoData Retriever, you can run it from inside R by doing things like:
install.packages('ecoretriever') library(ecoretriever) # List the datasets available via the Retriever ecoretriever::datasets() # Install the Gentry dataset into csv files in your working directory ecoretriever::install('Gentry', 'csv') # Download the raw Gentry dataset files, without any processing, # to the subdirectory named data ecoretriever::download('Gentry', './data/') # Install and load a dataset as a list Gentry = ecoretriever::fetch('Gentry') names(Gentry) head(Gentry$counts)
The other big advance in this release is the ability to have the Retriever directly download files instead of processing them. This allows us to support data that doesn’t come in standard tabular forms. So, we can now include things like environmental data in GIS formats and phylogenetic data such as supertrees. We’ve used this new capability to allow the automatic downloading of the Bioclim data, one of the most widely used climate datasets in ecology, and the supertree for mammals from Fritz et al. 2009.
Finally, we’ve also add the very cool mammalian diet dataset from Dryad
Ecology Letters now allows preprints; and why this is a big deal for ecology
As announced by Noam Ross on Twitter (and confirmed by the Editor in Chief of Ecology Letters), Ecology Letters will now allow the submission of manuscripts that have been posted as preprints. Details will be published in an editorial in Ecology Letters. I want to say a heartfelt thanks to Marcel Holyoak and the entire Ecology Letters editorial board for listening to the ecological community and modifying their policies. Science is working a little better today than it was yesterday thanks to their efforts.
For those of you who are new to the concept of preprints, they are manuscripts, that have not yet been published in peer reviewed journals, which are posted to websites like arXiv, PeerJ, and bioRxiv. This process allows for more rapid communication of scientific results and improved quality of published papers though more expansive pre-publication peer-review. If you’d like to read more check out our paper on The Case for Open Preprints in Biology.
The fact that Ecology Letters now allows preprints is a big deal for ecology because they were the last of the major ecology journals to make the transition. The ESA journals began allowing preprints just over two years ago and the BES journals made the switch about 9 months ago. In addition, Science, Nature, PNAS, PLOS Biology, and a number of other ecology journals (e.g., Biotropica) all support preprints. This means that all of the top ecology journals, and all of the top general science journals that most ecologists publish in, allow the posting of preprints. As such, there is not longer a reason to not post preprints based on the possibility of not being able to publish in a preferred journal. This can potentially shave months to years off of the time between discovery and initial communication of results in ecology.
It also means that other ecology journals that still do not allow the posting of preprints are under significant pressure to change their policies. With all of the big journals allowing preprints they have no reasonable excuse for not modernizing their policies, and they risk loosing out on papers that are initially submitted to higher profile journals and are posted as preprints.
It’s a good day for science. Celebrate by posting your next manuscript as a preprint.
Which preprint server should I use?
Preprints are rapidly becoming popular in biology as a way to speed up the process of science, get feedback on manuscripts prior to publication, and establish precedence (Desjardins-Proulx et al. 2013). Since biologists are still learning about preprints I regularly get asked which of the available preprint servers to use. Here’s the long-form version of my response.
The good news is that you can’t go wrong right now. The posting of a preprint and telling people about it is far more important than the particular preprint server you choose. All of the major preprint servers are good choices.Of course you still need to pick one and the best way to do that is to think about the differences between available options. Here’s my take on four of the major preprint servers: arXiv, bioRxiv, PeerJ, and figshare.
arXiv is the oldest of the science preprint servers. As a result it is the most well established, it is well respected, more people have heard of it than any of the other preprint servers, and there is no risk of it disappearing any time soon. The downside to having been around for a long time is that arXiv is currently missing some features that are increasingly valued on the modern web. In particular there is currently no ability to comment on preprints (though they are working on this) and there are no altmetrics (things like download counts that can indicate how popular a preprint is). The other thing to consider is that arXiv’s focus is on the quantitative sciences, which can be both a pro and a con. If you do math, physics, computer science, etc., this is the preprint server for you. If you do biology it depends on the kind of research you do. If your work is quantitative then your research may be seen by folks outside of your discipline working on related quantitative problems. If your work isn’t particularly quantitative it won’t fit in as well. arXiv allows an array of licenses that can either allow or restrict reuse. In my experience it can take about a week for a preprint to go up on arXiv and the submission process is probably the most difficult of the available options (but it’s still far easier than submitting a paper to a journal).
bioRxiv is the new kid on the block having launched less than a year ago. It has both commenting and altmetrics, but whether it will become as established as arXiv and stick around for a long time remains to be seen. It is explicitly biology focused and accepts research of any kind in the biological sciences. If you’re a biologist, this means that you’re less likely to reach people outside of biology, but it may be more likely that biology folks come across your work. bioRxiv allows an array of licenses that can either allow or restrict reuse. However, they explicitly override the less open licenses for text mining purposes, so all preprints there can be text-mined. In my experience it can take about a week for a preprint to go up on bioRxiv.
PeerJ Preprints is another new preprint server that is focused on biology and accepts research from across the biological sciences. Like bioRxiv it has commenting and altmetrics. It is the fastest of the preprint servers, with less than 24 hours from submission to posting in my experience. PeerJ has a strong commitment to open access, so all of it’s preprints are licensed with the Creative Commons Attribution License. PeerJ also publishes an open access journal, but you can post preprints to PeerJ Preprints with out submitting them to the journal (and this is very common). If you do decide to submit your manuscript to the PeerJ journal after posting it as a preprint you can do this with a single click and, should it be published, the preprint will be linked to the paper. PeerJ has the most modern infrastructure of any of the preprint servers, which makes for really pleasant submission, reading, and commenting experiences. You can also earn PeerJ reputation points for posting preprints and engaging in discussions about them. PeerJ is the only major preprint server run by a for-profit company. This is only an issue if you plan to submit your paper to a journal that only allows the posting of non-commercial preprints. I only know of only one journal with this restriction, but it is American Naturalist which can be an important journal in some areas of biology.
figshare is a place to put any kind of research output including data, figures, slides, and preprints. The benefit of this general approach to archiving research outputs is that you can use figshare to store all kinds of research outputs in the same place. The downside is that because it doesn’t focus on preprints people may be less likely to find your manuscript among all of the other research objects. One of the things I like about this broad approach to archiving anything is that I feel comfortable posting that isn’t really manuscripts. For example, I post grant proposals there. figshare accepts research from any branch of science and has commenting and altmetrics. There is no delay from submission to posting. Like PeerJ, figshare is a for-profit company and any document posted there will be licensed with the Creative Commons Attribution License.
Those are my thoughts. I have preprints on all three preprint servers + figshare and I’ve been happy with all three experiences. As I said at the beginning, the most important thing is to help speed up the scientific process by posting your work as preprints. Everything else is just details.
UPDATE: It looks like due to a hiccup with scheduling this post than an early version went out to some folks without the figshare section.
UPDATE: In the comments Richard Sever notes that bioRxiv’s preprints are typically posted within 48 hours of submission and that their interpretation of the text mining clause is that this is covered by fair use. See our discussion in the comments for more details.
Sharing in Science: my full reply to Eli Kintisch
A couple of weeks ago Eli Kintisch (@elikint) interviewed me for what turned out to be a great article on “Sharing in Science” for Science Careers. He also interviewed Titus Brown (@ctitusbrown) who has since posted the full text of his reply, so I thought I’d do the same thing.
How has sharing code, data, R methods helped you with your scientific research?
Definitely. Sharing code and data helps the scientific community make more rapid progress by avoiding duplicated effort and by facilitating more reproducible research. Working together in this way helps us tackle the big scientific questions and that’s why I got into science in the first place. More directly, sharing benefits my group’s research in a number of ways:
- Sharing code and data results in the community being more aware of the research you are doing and more appreciative of the contributions you are making to the field as a whole. This results in new collaborations, invitations to give seminars and write papers, and access to excellent students and postdocs who might not have heard about my lab otherwise.
- Developing code and data so that it can be shared saves us a lot of time. We reuse each others code and data within the lab for different projects, and when a reviewer requests a small change in an analysis we can make a small change in our code and then regenerate the results and figures for the project by running a single program. This also makes our research more reproducible and allows me to quickly answer questions about analyses years after they’ve been conducted when the student or postdoc leading the project is no longer in the lab. We invest a little more time up front, but it saves us a lot of time in the long run. Getting folks to work this way is difficult unless they know they are going to be sharing things publicly.
- One of the biggest benefits of sharing code and data is in competing for grants. Funding agencies want to know how the money they spend will benefit science as a whole, and being able to make a compelling case that you share your code and data, and that it is used by others in the community, is important for satisfying this goal of the funders. Most major funding agencies have now codified this requirement in the form of data management plans that describe how the data and code will be managed and when and how it will be shared. Having a well established track record in sharing makes a compelling argument that you will benefit science beyond your own publications, and I have definitely benefited from that in the grant review process.
What barriers exist in your mind to more people doing so?
There is a lot of fear about openly sharing data and code. People believe that making their work public will result in being scooped or that their efforts will be criticized because they are too messy. There is a strong perception that sharing code and data takes a lot of extra time and effort. So the biggest barriers are sociological at the moment.
To address these barriers we need to be a better job of providing credit to scientists for sharing good data and code. We also need to do a better job of educating folks about the benefits of doing so. For example, in my experience, the time and effort dedicated to developing and documenting code and data as if you plan to share it actually ends up saving the individual research time in the long run. This happens because when you return to a project a few months or years after the original data collection or code development, it is much easier if the code and data are in a form that makes it easy to work with.
How has twitter helped your research efforts?
Twitter has been great for finding out about exciting new research, spreading the word about our research, getting feedback from a broad array of folks in the science and tech community, and developing new collaborations. A recent paper that I co-authored in PLOS Biology actually started as a conversation on twitter.
How has R Open Science helped you with your work, or why is it important or not?
rOpenSci is making it easier for scientists to acquire and analyze the large amounts of scientific data that are available on the web. They have been wrapping many of the major science related APIs in R, which makes these rich data sources available to large numbers of scientists who don’t even know what an API is. It also makes it easier for scientists with more developed computational skills to get research done. Instead of spending time figuring out the APIs for potentially dozens of different data sources, they can simply access rOpenSci’s suite of packages to quickly and easily download the data they need and get back to doing science. My research group has used some of their packages to access data in this way and we are in the process of developing a package with them that makes one of our Python tools for acquiring ecological data (the EcoData Retriever) easy to use in R.
Any practical tips you’d share on making sharing easier?
One of the things I think is most important when sharing both code and data is to use standard licences. Scientists have a habit of thinking they are lawyers and writing their own licenses and data use agreements that govern how the data and code and can used. This leads to a lot of ambiguity and difficulty in using data and code from multiple sources. Using standard open source and open data licences vastly simplifies the the process of making your work available and will allow science to benefit the most from your efforts.
And do you think sharing data/methods will help you get tenure? Evidence it has helped others?
I have tenure and I certainly emphasized my open science efforts in my packet. One of the big emphases in tenure packets is demonstrating the impact of your research, and showing that other people are using your data and code is a strong way to do this. Whether or not this directly impacted the decision to give me tenure I don’t know. Sharing data and code is definitely beneficial to competing for grants (as I described above) and increasingly to publishing papers as many journals now require the inclusion of data and code for replication. It also benefits your reputation (as I described above). Since tenure at most research universities is largely a combination of papers, grants, and reputation, and I think that sharing at least increases one’s chances of getting tenure indirectly.
UPDATE: Added missing link to Titus Brown’s post: http://ivory.idyll.org/blog/2014-eli-conversation.html
EcoData Retriever: quickly download and cleanup ecological data so you can get back to doing science
If you’ve every worked with scientific data, your own or someone elses, you know that you can end up spending a lot of time just cleaning up the data and getting it in a state that makes it ready for analysis. This involves everything from cleaning up non-standard nulls values to completely restructuring the data so that tools like R, Python, and database management systems (e.g., MS Access, PostgreSQL) know how to work with them. Doing this for one dataset can be a lot of work and if you work with a number of different databases like I do the time and energy can really take away from the time you have to actually do science.
Over the last few years Ben Morris and I been working on a project called the EcoData Retriever to make this process easier and more repeatable for ecologists. With a click of a button, or a single call from the command line, the Retriever will download an ecological dataset, clean it up, restructure and assemble it (if necessary) and install it into your database management system of choice (including MS Access, PostgreSQL, MySQL, or SQLite) or provide you with CSV files to load into R, Python, or Excel.
Just click on the box to get the data:
Or run a command like this from the command line:
retriever install msaccess BBS --file myaccessdb.accdb
This means that instead of spending a couple of days wrangling a large dataset like the North American Breeding Bird Survey into a state where you can do some science, you just ask the Retriever to take care of it for you. If you work actively with Breeding Bird Survey data and you always like to use the most up to date version with the newest data and the latest error corrections, this can save you a couple of days a year. If you also work with some of the other complicated ecological datasets like Forest Inventory and Analysis and Alwyn Gentry’s Forest Transect data, the time savings can easily be a week.
The Retriever handles things like:
- Creating the underlying database structures
- Automatically determining delimiters and data types
- Downloading the data (and if there are over 100 data files that can be a lot of clicks)
- Transforming data into standard structures so that common tools in R and Python and relational database management systems know how to work with it (e.g., converting cross-tabulated data)
- Converting non-standard null values (e.g., 999.0, -999, NoData) into standard ones
- Combining multiple data files into single tables
- Placing all related tables in a single database or schema
The EcoData Retriever currently includes a number of large, openly available, ecological datasets (see a full list here). It’s also easy to add new datasets to the EcoData Retriever if you want to. For simple data tables a Retriever script can be as simple as:
name: Name of the dataset description: A brief description of the dataset of ~25 words. shortname: A one word name for the dataset table: MyTableName, http://awesomedatasource.com/dataset
The Retriever has an installer for Windows, an App for Mac, and a package for Ubuntu/Debian Linux. See the quick explanation of how to get started and then go take it for a spin.
If you’re interested in reading more about the Retriever you can checkout the website or read our paper on the project.
We also have some exciting new features on the To Do list including:
- Automatically cleaning up the taxonomy using existing services
- Providing detailed tracking of the provenance of your data by recording the date it was downloaded, the version of the software used, and information about what cleanup steps the Retriever performed
- Integration into R and Python
Let us know what you think we should work on next in the comments.
British Ecological Society journals now allow preprints
The British Ecological Society has announced that will now allow the submission of papers with preprints (formal language here). This means that you can now submit preprinted papers to Journal of Ecology, Journal of Animal Ecology, Methods in Ecology and Evolution, Journal of Applied Ecology, and Functional Ecology. By allowing preprints BES joins the Ecological Society of America which instituted a pro-preprint policy last year. While BES’s formal policy is still a little more vague than I would like*, they have confirmed via Twitter that even preprints with open licenses are OK as long as they are not updated following peer review.
Preprints are important because they:
- Speed up the progress of science by allowing research to be discussed and built on as soon as it is finished
- Allow early career scientists to establish themselves more rapidly
- Improve the quality of published research by allowing a potentially large pool reviewers to comment on and improve the manuscript (see our excellent experience with this)
BES getting on board with preprints is particularly great news because the number of ecology journals that do not allow preprints is rapidly shrinking to the point that ecologists will no longer need to consider where they might want to submit their papers when deciding whether or not to post preprints. The only major blocker at this point to my mind is Ecology Letters. So, my thanks to BES for helping move science forward!
*Which is why I waited 3 weeks for clarification before posting.
Exploring MaxEnt based species-area relationship predictions [Research Summary]
This is a guest post by Dan McGlinn, a weecology postdoc (@DanMcGlinn on Twitter). It is a Research Summary of: McGlinn, D.J., X. Xiao, and E.P. White. 2013. An empirical evaluation of four variants of a universal species–area relationship. PeerJ 1:e212 http://dx.doi.org/10.7717/peerj.212. These posts are intended to help communicate our research to folks who might not have the time, energy, expertise, or inclination to read the full paper, but who are interested in a <1000 general language summary.
It is well established in ecology that if the area of a sample is increased you will in general see an increase in the number species observed. There are a lot of different reasons why larger areas harbor more species: larger areas contain more individuals, habitats, and environmental variation, and they are likely to cross more barriers to dispersal – all things that promote more species to be able to exist together in an area. We typically observe relatively smooth and simple looking increases in species number with area. This observation has mystified ecologists: How can a pattern that should be influenced by many different and biologically idiosyncratic processes appear so similar across scales, taxonomic groups, and ecological systems?
Recently a theory was proposed (Harte et al. 2008, Harte et al. 2009) which suggests that detailed knowledge of the complex processes that influence the increase in species number may not be necessary to accurately predict the pattern. The theory proposes that ecological systems tend to simply be in their most likely configuration. Specifically, the theory suggests that if we have information on the total number of species and individuals in an area then we can predict the number of species in smaller portions of that area.
Published work on this new theory suggests that it has potential for accurately predicting how species number changes with area; however, it has not been appreciated that there are actually four different ways that the theory can be operationalized to make a prediction. We were interested to learn
- Can the theory accurately predict how species number changes with area across many different ecological systems, and
- Do the different versions of the theory consistently perform better than others
To answer these questions we needed data. We searched online and made requests to our colleagues for datasets that documented the spatial configuration of ecological communities. We were able to pull together a collection of 16 plant community datasets. The communities spanned a wide range of systems including hyper-diverse, old-growth tropical forests, a disturbance prone tropical forest, temperate oak-hickory and pine forests, a Mediterranean mixed-evergreen forest, a low diversity oak woodland, and a serpentine grassland.
Fig 1. A) Results from one of the datasets, the open circles display the observed data and the lines are the four different versions of the theory we examined. B) A comparison of the observed and predicted number of species across all areas and communities we examined for one of the versions of the theory.
Across the different communities we found that the theory was generally quite accurate at predicting the number of species (Fig 1 above), and that one of the versions of the theory was typically better than the others in terms of the accuracy of its predictions and the quantity of information it required to make predictions. There were a couple of noteworthy exceptions in our results. The low diversity oak woodland and the serpentine grassland both displayed unusual patterns of change in richness. The species in the serpentine grassland were more spatially clustered than was typically observed in the other communities and thus better described by the versions of the theory that predicted stronger clustering. Abundance in the oak woodland was primarily distributed across two species whereas the other 5 species where only observed once or twice. This unusual pattern of abundance resulted in a rather unique S-shaped relationship between the number of species and area and required inputting the observed species abundances to accurately model the pattern.
The two key findings from our study were
- The theory provides a practical tool for accurately predicting the number of species in sub-samples of a given site using only information on the total number of species and individuals in that entire area.
- The different versions of the theory do make different predictions and one appears to be superior
Of course there are still a lot of interesting questions to address. One question we are interested in is whether or not we can predict the inputs of the theory (total number of species and individuals for a community) using a statistical model and then plug those predictions into the theory to generate accurate fine-scaled predictions. This kind of application would be important for conservation applications because it would allow scientists to estimate the spatial pattern of rarity and diversity in the community without having to sample it directly. We are also interested in future development of the theory that provides predictions for the number of species at areas that are larger (rather than smaller) than the reference point which may have greater applicability to conservation work.
The accuracy of the theory also has the potential to help us understand the role of specific biological processes in shaping the relationship between species number and area. Because the theory didn’t include any explicit biological processes, our findings suggest that specific processes may only influence the observed relationship indirectly through the total number of species and individuals. Our results do not suggest that biological processes are not shaping the relationship but only that their influence may be rather indirect. This may be welcome news to practitioners who rely on the relationship between species number and area to devise reserve designs and predict the effects of habitat loss on diversity.
If you want to learn more you can read the full paper (it’s open access!) or check out the code underlying the analysis (it’s open source and includes instructions for replicating the analysis!).
Harte, J., A. B. Smith, and D. Storch. 2009. Biodiversity scales from plots to biomes with a universal species-area curve. Ecology Letters 12:789–797.
Harte, J., T. Zillio, E. Conlisk, and A. B. Smith. 2008. Maximum entropy and the state-variable approach to macroecology. Ecology 89:2700–2711.