We discussed whether we should be using GitLab or GitHub. Main advantages of GitLab are that it can be 'internally' shared as well as 'externally', so we can share things within WIN before flipping them to be externally visible. Straightforward to clone a repository from GitLab to GitHub when required.
We'll all try to start up a project in the GitLab open imaging site by Wendesday November 8th, when Duncan is presenting it to the WIN. If it turns out there are too many teething problems during this time, easy to switch to GitHub.
People already using GitLab/GitHub for sharing can just clone an existing project onto the site.
We’d like to have this in place for Wednesday 8th November: i.e. when Duncan is presenting GitLab to the FMRIB Wednesday meeting
We then discussed packages for stimulus presentation, with input from Chris Gallagher and Dave Flitney
Major requirements: interfacing with scanner via parallel port, accurate timings, interfacing with various button boxes etc., ability to run jointly on scanner and web
Presentation (NeuroBS) is currently only package that is guaranteed support within FMRIB. Advantages are reliable timing, can be programmed straightforwardly independently of scanner, and has good backwards compatibility. Downsides are licensing (bad for sharing), separate language from one you might be learning otherwise (i.e. MATLAB/Python), unable to run things on web, certain experiments can't be run.
Psychtoolbox, MATLAB-based, is run by a number of researchers on laptops. Although it can produce accurate timings if well controlled by experimenter, there have been problems with timings in the past, and with experiments not being translated across future versions of MATLAB as it gets updated. Chris and Dave were happy for people to continue to use it, but didn't think it was a good longer-term solution, particularly as more people are migrating to Python anyway.
PsychoPy, Python-based, now has funding from Wellcome Trust for full-time developer to support professionally. Being Python-based means that people will be able to learn it alongside coding analysis etc. in Python. It has a separate builder mode, helpful for people who are scared of coding, or who want to see what code for an experiment 'should' look like. It will soon have support to compile these experiments into PsychoJS, which will allow experiments to be run on Web. Being open-access means it is best for sharing.
Our recommendation is therefore for people to learn PsychoPy to code future experiments. This will become supported within WIN. We'll also maintain compatibility with Presentation for the foreseeable future, and people who want to stick to MATLAB can carry on using Psychtoolbox, as needed.
Separately, we discussed a bit about the back-end required for hosting online experiments, and whether further support could be provided for this within WIN. LH will look into this further.
Our plan for next month is a talk from Verena Heise on Open Publishing, and ‘How to get from Open Science to Robust Science’. Friday 6th October, Big Data Institute