Alan Turing Institute — Data Visualisation Workshops
Megan Fitzsimons and Kevin Donkers co-wrote this blog post following attendance at the Alan Turing Institute Data Visualisation Workshop.
On September 13th, we headed to the British Library in London to take part in a day of Data Visualisation workshops at the Alan Turing Institute. Data Visualisation has long been at the core of what we do, both within the Informatics Lab and in the Met Office as a whole. We were therefore interested to hear about some of the cutting edge projects emerging from leading data visualisation research. Fortunately, the workshop didn’t disappoint. During an afternoon of quick-fire talks, we were presented with some exciting implementations of data visualisation techniques, particularly in fields not considered ‘traditional’ for the subject. What follows is a quick summary of our thoughts from the day.
Tableau: Does it live up to the hype?
First on the workshop bill at Turing was an introductory tutorial to the Tableau, a “business intelligence software that helps people see and understand their data.” Although an impressive piece of software, our primary takeaway was that it perhaps wasn’t suited to our needs. Here’s why:
Pros:
- In terms of accessibility, Tableau can be used by pretty much anyone. Visualisations can be made to be as simple or complicated as necessary, which means the user experience can be anything from click and drag, to writing scripts in R or Python, to importing MatLab functions.
- Possibly the best feature of the Tableau software is the design of data storage and loading. When working with relatively hefty datasets, only the metadata is loaded in the first instance. Tableau then loads ‘extracts’, or in other words, loads only the chunk of data you need to work with. This practice makes the software extremely responsive, with very little lag.
- Collaboration is facilitated via packaging systems for documents. Exporting files works similarly to Adobe project packaging, where everything is wrapped in a folder to ensure there are no missing components when opened by the recipient. I was also impressed that the software considers and adheres to data protection, with different file formats to choose from depending on the level of privacy you need to retain around the data.
Cons:
- Put simply, it’s expensive. You’re also tied into complicated long term contracts, with a hefty amount of telemarketing (two calls within the first week of downloading in my case).
- As far as we could tell, supported data formats are limited. We work mostly with gridded geospatial data, which caused no end of difficulties when trying to import. I recognise that this wouldn’t be an issue for the majority of users, but it makes it unsuitable for us.
- I’m under the impression that the software is fairly opinionated in terms of working process, data preparation, and visualisation formats. If you’re willing to follow the suggested working process, I have no doubt that Tableau would work just fine. If you want to experiment and make out-of-the-ordinary visualisations, it might not be the right software for the project.
The latest in Data Visualisation research
The highlight of the day was the collection of presentations on visualisation projects happening across the UK. It’s particularly good to see that visualisation is being integrated into a wide variety of different topics, and is being recognised as an important tool for data analysis and communication. Amongst the many interesting talks, we were particularly impressed by project Quill, a framework for modeling negotiated texts in an open and collaborative way. Data provenance, tagging and reusability is something we’ve spent some time exploring recently, and it appears that the team behind the project have worked toward solving a few of these shared problems.
Other areas being explored included:
- Visualisation of missing data points and the information stored within. This particular example from Dr. Sara Fernstad showed how missing data in patient surveys could be used to show trends of their own. Dr Fernstad also demonstrated the need for care when deciding what to do with missing data, given that removal of imputation can significantly impact overall trends in the data.
- Dr. Hamish Carr introduced topological analyses of datasets as methods for pulling out key patterns and features. This is particularly useful for large, multidimensional datasets that are infeasible to visualise with traditional methods or contain features that are missed using these techniques.
- Prof. Nick Holliman presented findings from the ambitious Newcastle Urban Observatory project which used virtual reality to communicate data about the city of Newcastle to citizens, planners and emergency responders. Of particular interest was the method of displaying uncertainty using a visual representation of entropy. This was of particular interest to emergency responders who had to ingest a large amount of information with varying levels of uncertainty in order to make decisions.
Overall, I think our main takeaway from the day was acknowledging the value gained from communities of interest, such as that for data visualisation. Throughout the course of the day, it was interesting to identify areas of shared difficulties. Hearing how other organisations have overcome some of the challenges we’ve identified is a great development tool for our own research.