diff --git a/images/nbgrader.jpg b/images/nbgrader.jpg new file mode 100644 index 0000000..171d620 Binary files /dev/null and b/images/nbgrader.jpg differ diff --git a/posts/2017-jul14.html b/posts/2017-jul14.html new file mode 100644 index 0000000..13055b4 --- /dev/null +++ b/posts/2017-jul14.html @@ -0,0 +1,40 @@ + + + + + + + + +

See original posting here.

+ +

I just got back from SciPy2017 (I had a talk on ReproZip accepted - slides)and I learned about some amazing open source tools for research! This year, SciPy 2017 was in Austin, Texas from July 10-16, 2017. It was the 16th annual Scientific Computing with Python Conference, and focused on great new tools and methods for research with Python.

+ +

These are my top 5 favourite takeaways from SciPy 2017!

+ + + +
    +
  1. SciSheets: Anyone who knows me knows that I really can't stand Excel. It encodes your data weirdly, and is such a black box it causes more errors in research than it ever helps analysis. This is why I was pumped to see a session on building a better spreadsheet - one that combines programming with the simplicity of spreadsheets. SciSheets is a web application that allows users to run Python expressions or scripts in a spreadhseet, but also export spreadsheets to a standalone Python program! You can find a demo video here!
    2. +
  2. nbgrader: This is a phenomenal application for assignment management and grading in Jupyter notebooks. The nbgrader extension for Jupyter notebooks guides the instructor through assignment and grading tasks using the familiar Jupyter notebook interface. It's made up of a few Jupyter Notebook extensions. The formgrader extension allows instructors to use functionality from nbgrader to generate student versions of assignments (including releasing to students), collecting assignments, and auto and manual grading submissions. Students just work in the notebook and submit! You can read more at the GitHub repo.
    3. + +
    +

    The nbgrader workflows from the SciPy presentation.

    + +
  3. Dataflow: This extension to Jupyter Notebooks answers the question, "how can a notebook be structured so rewriting isn't necessary?" and "how can cells in a notebook be linked more robustly?" Their solution was to make cell IDs persistent, similarly to UUIDs. This allows users to powerfully reference previous outputs. You see the slides from SciPy here.
    4. +
  4. The Journal of Open Source Software: Ok, I didn't just learn about JOSS (I have a paper there!) but it's still one of my favourite things. It's an open source journal for software. Developers just have to write a short essay (2 paragraph markdown file with some references and an image) and have their code available for review on GitHub. The reviews look at the source code, and test it out before acceptance. From their website: "The Journal of Open Source Software (JOSS) is an academic journal with a formal peer review process that is designed to improve the quality of the software submitted." It's a great way for developers in academia to get their work reviewed, and get credit for their excellent software.
    5. + +
  5. Elegant SciPy book: Written by Juan Nunez-Iglesias (@jni), Harriet Dashnow (@hdashnow), and Stéfan van der Walt (@stefanv), and published by O'Reilly Media, this fully free and open book focuses on the foundations of scientific python. You can download the book from the GitHub repository as Markdown or an executable Jupyter Notebook. Great work done on opening the book in a machine readable and executable format!!
    6. +
+ + + diff --git a/posts/2017-may26.html b/posts/2017-may26.html new file mode 100644 index 0000000..75c78d6 --- /dev/null +++ b/posts/2017-may26.html @@ -0,0 +1,46 @@ + + + + + + + + +

See original posting here.

+

I just got back from IASSIST 2017 and I have to say...I was very impressed! This year, IASSIST (The International Association for Social Science Information Services & Technology) 2017 was in Lawrence, Kansas from May 23-26, 2017. True to it's name, this conference brought people from all around the world:

+ +
+

A map of #iassist17 attendees! pic.twitter.com/V6fV5Ey5iv

+ — Vicky Steeves (@VickySteeves) May 24, 2017
+ + +

These are my top 5 favourite takeaways from IASSIST 2017:

+ +
    +
  1. An interesting project that was recently published in PLoS One, Research data management in academic institutions: A scoping review, which was presented as a poster during the conference. This was essentially a systematic review that was designed to describe the volume, topics, and methodologies of existing scholarly work on research data management in academia. They looked at 301 articles out of the original 13,002 titles. They made the data (the text, methods, etc.) available on Zenodo: Dataset for: Research data management in academic institutions: a scoping review!
    2. +
  2. Packrat: a dependency manager in R that looks to solve the problem of "dependency hell" -- that software depends on other packages to run, and these change all the time with no warning, and these changes can break existing code. Packrat works by making a project specific package library, rather than using R's native package manager (which updates libraries as they are released). This means the R code can be packaged up with its dependencies. However, it doesn't pack the version of R, which can pose problems.
    3. +
  3. Sam Spencer of the Aristotle metadata registry gave a great talk about work done in the open metadata space, giving a strong usecase: government data hosted on data.gov.au. He shocked the crowd by keeping metadata in CSV format. He asks for 10 basic fields of metadata from users in CSV form -- and there it stays! He mentioned he was scared to admit this to this crowd, but it's yielded good things for him, including data linkages without explicitly doing linked data. He spoke specifically about using this for geo-metadata; you can check out how it's worked out on this map.
    4. +
  4. One of the more interesting talks I went to was about digital preservation of 3D data! The speaker laid out 5 methods of creation: freeform (like CAD), measurement, observation, "mix," and algorithm/scanning or photogrammetry. 3D data is difficult to preserve mainly because of a lack of standards, particularly metadata standards. The speaker presented a case study that used Dublin Core as a basis for metadata for the Awash National Park Baboon Research Project's 3D data.
    5. +
  5. The Digital Curation Network gave an update on their initial planning grant. The DCN allows universities to staff share on data curation, which often is too much for one data curator. The first grant allowed six universities to test how local curation practices translates into a network practice. The next phase includes implementation of the network, during which time other institutions can join. The network also came out with centralized steps for curation: +
      +
    1. Check data files and read documentation
      2. +
    2. Understand/try to understand the data
      3. +
    3. Request missing information or changes
      4. +
    4. Augment the submission with metadata
      5. +
    5. Transform file format for reuse and long-term preservation
      6. +
    6. Evaluate and rate the overall submission using FAIR 
      7. +
    +
    6. +
+ + +