d3.js crossed my path a couple of times yesterday: firstly, in the form of an enquiry about whether I’d be interested in writing a book on d3.js (I’m not sure I’m qualified: as I responded, I’m more of a script kiddie who sees things I can reuse, rather than have any understanding at all about how d3.js does what it does…); secondly, via a link to d3.js creator Mike Bostock’s new demo of Sankey diagrams built using d3.js:

Hmm… Sankey diagrams are good for visualising flow, so to get to grips myself with seeing if I could plug-and-play with the component, I needed an appropriate data set. F1 related data is usually my first thought as far as testbed data goes (no confidences to break, the STEM/innovation outreach/tech transfer context, etc etc) so what things flow in F1? What quantities are conserved whilst being passed between different classes of entity? How about points… points are awarded on a per race basis to drivers who are members of teams. It’s also a championship sport, run over several races. The individual Driver Championship is a competition between drivers to accumulate the most points over the course of the season, and the Constructor Chanmpionship is a battle between teams. Which suggests to me that a Sankey plot of points from races to drivers and then constructors might work?

So what do we need to do? First up, look at the source code for the demo using View Source. Here’s the relevant bit:

Data is being pulled in from a relatively addressed file, energy.json. Let’s see what it looks like:

Okay – a node list and an edge list. From previous experience, I know that there is a d3.js JSON exporter built into the Python networkx library, so maybe we can generate the data file from a network representation of the data in networkx?

Here we are: node_link_data(G) “[r]eturn data in node-link format that is suitable for JSON serialization and use in Javascript documents.”

Next step – getting the data. I’ve already done a demo of visualising F1 championship points sourced from the Ergast motor racing API as a treemap (but not blogged it? Hmmm…. must fix that) that draws on a JSON data feed constructed from data extracted from the Ergast API so I can clone that code and use it as the basis for constructing a directed graph that represents points allocations: race nodes are linked to driver nodes with edges weighted by points scored in that race, and driver nodes are connected to teams by edges weighted according to the total number of points the driver has earned so far. (Hmm, that gives me an idea for a better way of coding the weight for that edge…)

I don’t have time to blog the how to of the code right now – train and boat to catch – but will do so later. If you want to look at the code, it’s here: Ergast Championship nodelist. And here’s the result – F1 Chanpionship 2012 Points as a Sankey Diagram:

See what I mean about being a cut and paste script kiddie?!;-)

A post on the Guardian Datablog yesterday (Higher education funding: which institutions will be affected?) alerted me to the release of HEFCE’s “provisional allocations of recurrent funding for teaching and research, and the setting of student number control limits for institutions, for academic year 2012-13″ (funding data).

Here are the OU figures for teaching:

HEFCE preliminary teaching funding allocations to the Open University, 2012-13

Of the research funding for 2012-13, mainstream funding was 8,030,807, the RDP supervision fund came in at 1,282,371, along with 604,103 “other”, making up the full 9,917,281 research allocation.

Adding Higher Education Innovation Funding of 950,000, the OU’s total allocation was 139,714,060.

So what other funding comes into the universities from public funds?

Open Spending publishes data relating to spend by government departments to named organisations, so we can search that for data spent by government departments with the universities (for example, here is a search on OpenSpending.org for “open university”:

Given the amounts spent by public bodies on consultancy (try searching OpenCorporates for mentions of PriceWaterhouseCoopers, or any of EDS, Capita, Accenture, Deloitte, McKinsey, BT’s consulting arm, IBM, Booz Allen, PA, KPMG (h/t @loveitloveit)), university based consultancy may come in reasonably cheaply?

The universities also receive funding for research via the UK research councils (EPSRC, ESRC, AHRC, MRC, BBSRC, NERC, STFC) along with innovation funding from JISC. Unpicking the research council funding awards to universities can be a bit of a chore, but scrapers are appearing on Scraperwiki that make for easier access to individual grant awards data:

• AHRC funding scraper; [grab data using queries of the form select * from `swdata` where organisation like "%open university%" on scraper arts-humanities-research-council-grants]
• EPSRC funding scraper; [grab data using queries of the form select * from `grants` where department_id in (select distinct id as department_id from `departments` where organisation_id in (select id from `organisations` where name like "%open university%")) on scraper epsrc_grants_1]
• ESRC funding scraper; [grab data using queries of the form select * from `grantdata` where institution like "%open university%" on scraper esrc_research_grants]
• BBSRC funding [broken?] scraper;
• NERC funding [broken?] scraper;
• STFC funding scraper; [grab data using queries of the form select * from `swdata` where institution like "%open university%" on scraper stfc-institution-data]

In order to get a unified view over the detailed funding of the institutions from these different sources, the data needs to be reconciled. There are several ID schemes for identifying universities (eg UCAS or HESA codes; see for example GetTheData: Universities by Mission Group) but even official data releases tend not make use of these, preferring instead to rely solely on insitution names, as for example in the case of the recent HEFCE provisional funding data release [DOh! This is not the case – identifiers are there, apparently (I have to admit, I didn’t check and was being a little hasty… See the contribution/correction from David Kernohan in the comments to this post…]:

For some time, I’ve been trying to put my finger on why data releases like this are so hard to work with, and I think I’ve twigged it… even when released in a spreadsheet form, the data often still isn’t immediately “database-ready” data. Getting data from a spreadsheet into a database often requires an element of hands-on crafting – coping with rows that contain irregular comment data, as well as handling columns or rows with multicolumn and multirow labels. So here are a couple of things that would make life easier in the short term, though they maybe don’t represent best practice in the longer term…:

1) release data as simple CSV files (odd as it may seem), because these can be easily loaded into applications that can actually work on the data as data. (I haven’t started to think too much yet about pragmatic ways of dealing with spreadsheets where cell values are generated by formulae, because they provide an audit trail from one data set to derived views generated from that data.)

2) have a column containing regular identifiers using a known identification scheme, for example, HESA or UCAS codes for HEIs. If the data set is a bit messy, and you can only partially fill the ID column, then only partially fill it; it’ll make life easier joining those rows at least to other related datasets…

As far as UK HE goes, the JISC monitoring unit/JISCMU has a an api over various administrative data elements relating to UK HEIs (eg GetTheData: Postcode data for HE and FE institutes, but I don’t think it offers a Google Refine reconciliation service, (ideally with some sort of optional string similarity service)…? Yet?! 😉 maybe that’d make for a good rapid innovation project???

PS I’m reminded of a couple of related things: Test Your RESTful API With YQL, a corollary to the idea that you can check your data at least works by trying to use it (eg generate a simple chart from it) mapped to the world of APIs: if you can’t easily generate a YQL table/wrapper for it, it’s maybe not that easy to use? 2) the scraperwiki/okf post from @frabcus and @rufuspollock on the need for data management systems not content management systems.

PPS Looking at the actual Guardian figures reveals all sorts of market levers appearing… Via @dkernohan, FT: A quiet Big Bang in universities

Secretary of State for Education, Michael Gove, gave a speech yesterday on rethinking the ICT curriculum in UK schools. You can read a copy of the speech variously on the Department for Education website, or, err, on the Guardian website.

Seeing these two copies of what is apparently the same speech, I started wondering:

a) which is the “best” source to reference?
b) how come the Guardian doesn’t add a disclaimer about the provenance of, and link, to the DfE version? [Note the disclaimer in the DfE version – “Please note that the text below may not always reflect the exact words used by the speaker.”]
c) is the Guardian version an actual transcript, maybe? That is, does the Guardian reprint the “exact words” used by the speaker?

And that made me think I should do a diff… About which, more below…

Before that, however, here’s a quick piece of reflection on how these two things – the reinvention of the the IT curriculum, and the provenance of, and value added to, content published on news and tech industry blog sites – collide in my mind…

So for example, I’ve been pondering what the role of journalism is, lately, in part because I’m trying to clarify in my own mind what I think the practice and role of data journalism are (maybe I should apply for a Nieman-Berkman Fellowship in Journalism Innovation to work on this properly?!). It seems to me that “communication” is one important part (raising awareness of particular issues, events, or decisions), and holding governments and companies to account is another. (Actually, I think Paul Bradshaw has called me out on that, before, suggesting it was more to do with providing an evidence base through verification and triangulation, as well as comment, against which governments and companies could be held to account (err, I think? As an unjournalist, I don’t have notes or a verbatim quote against which to check that statement, and I’m too lazy to email/DM/phone Paul to clarify what he may or may not have said…(The extent of my checking is typically limited to what I can find on the web or in personal archives…which appear to be lacking on this point…))

Another thing I’ve been mulling over recently in a couple of contexts relates to the notion of what are variously referred to as digital or information skills.

The first context is “data journalism”, and the extent to which data journalists need to be able to do programming (in the sense of identifying the steps in a process that can be automated and how they should be sequenced or organised) versus writing code. (I can’t write code for toffee, but I can read it well enough to copy, paste and change bits that other people have written. That is, I can appropriate and reuse other people’s code, but can’t write it from scratch very well… Partly because I can’t ever remember the syntax and low level function names. I can also use tools such as Yahoo Pipes and Google Refine to do coding like things…) Then there’s the question of what to call things like URL hacking or (search engine) query building?

The second context is geeky computer techie stuff in schools, the sort of thing covered by Michael Gove’s speech at the BETT show on the national ICT curriculum (or lack thereof), and about which the educational digerati were all over on Twitter yesterday. Over the weekend, houseclearing my way through various “archives”, I came across all manner of press clippings from 2000-2005 or so about the activities of the OU Robotics Outreach Group, of which I was a co-founder (the web presence has only recently been shut down, in part because of the retirement of the sys admin on whose server the websites resided.) This group ran an annual open meeting every November for several years hosting talks from the educational robotics community in the UK (from primary school to HE level). The group also co-ordinated the RoboCup Junior competition in the UK, ran outreach events, developed various support materials and activities for use with Lego Mindstorms, and led the EPSRC/AHRC Creative Robotics Research Network.

At every robotics event, we’d try to involve kids and/or adults in elements of problem solving, mechanical design, programming (not really coding…) based around some sort of themed challenge: a robot fashion show, for example, or a treasure hunt (both variants on edge following/line following;-) Or a robot rescue mission, as used in a day long activity in the “Engineering: An Active Introduction” (TXR120) OU residential school, or the 3 hour “Robot Theme Park” team building activity in the Masters level “Team Engineering” (T885) weekend school. [If you’re interested, we may be able to take bookings to run these events at your institution. We can make them work at a variety of difficulty levels from KS3-4 and up;-)]

Given that working at the bits-atoms interface is where the a lot of the not-purely-theoretical-or-hardcore-engineering innovation and application development is likely to take place over the next few years, any mandate to drop the “boring” Windows training ICT stuff in favour of programming (which I suspect can be taught in not only a really tedious way, but a really confusing and badly delivered way too) is probably Not the Best Plan.

Slightly better, and something that I know is currently being mooted for reigniting interest in computing, is the Raspberry Pi, a cheap, self-contained, programmable computer on a board (good for British industry, just like the BBC Micro was…;-) that allows you to work at the interface between the real world of atoms and the virtual world of bits that exists inside the computer. (See also things like the OU Senseboard, as used on the OU course “My Digital Life” (TU100).)

If schools were actually being encouraged to make a financial investment on a par with the level of investment around the introduction of the BBC Micro, back in the day, I’d suggest a 3D printer would have more of the wow factor…(I’ll doodle more on the rationale behind this in another post…) The financial climate may not allow for that (but I bet budget will manage to get spent anyway…) but whatever the case, I think Gove needs to be wary about consigning kids to lessons of coding hell. And maybe take a look at programming in a wider creative context, such as robotics (the word “robotics” is one of the reason why I think it’s seen as a very specialised, niche subject; we need a better phrase, such as “Creative Technologies”, which could combine elements of robotics, games programming, photoshop, and, yex, Powerpoint too… Hmm… thinks.. the OU has a couple of courses that have just come to the end of their life that between them provide a couple of hundred hours of content and activity on robotics (T184) and games programming (T151), and that we delivered, in part, to 6th formers under the OU’s Young Applicants in Schools Scheme.

Anyway, that’s all as maybe… Because there are plenty of digital skills that let you do coding like things without having to write code. Such as finding out whether there are any differences between the text in the DfE copy of Gove’s BETT speech, and the Guardian copy.

Copy the text from each page into a separate text file, and save it. (You’ll need a text editor for that..) Then, if you haven’t already got one, find yourself a good text editor. I use Text Wrangler on a Mac. (Actually, I think MS Word may have a diff function?)

The difference’s all tend to be in the characters used for quotation marks (character encodings are one of the things that can make all sorts of programmes fall over, or misbehave. Just being aware that they may cause a problem, as well as how and why, would be a great step in improving the baseline level understanding of folk IT. Some of the line breaks don’t quite match up either, but other than that, the text is the same.

Now, this may be because Gove was a good little minister and read out the words exactly as they had been prepared. Or it may be the case that the Guardian just reprinted the speech without mentioning provenance, or the disclaimer that he may not actually have read the words of that speech (I have vague memories of an episode of Yes, Minister, here…;-)

Whatever the case, if you know: a) that it’s even possible to compare two documents to see if they are different (a handy piece of folk IT knowledge); and b) know a tool that does it (or how to find a tool that does it, or a person that may have a tool that can do it), then you can compare the texts for yourself. And along the way, maybe learn that churnalism, in a variety of forms, is endemic in the media. Or maybe just demonstrate to yourself when the media is acting in a purely comms, rather than journalistic, role?

PS other phrases in the area: “computational thinking”. Hear, for example: A conversation with Jeannette Wing about computational thinking

PPS I just remembered – there’s a data journalism hook around this story too… from a tweet exchange last night that I was reminded of by an RT:

josiefraser: RT @grmcall: Of the 28,000 new teachers last year in the UK, 3 had a computer-related degree. Not 3000, just 3.
dlivingstone: @josiefraser Source??? Not found it yet. RT @grmcall: 28000 new UK teachers last year, 3 had a computer-related degree. Not 3000, just 3
josiefraser: That ICT qualification teacher stat RT @grmcall: Source is the Guardian http://www.guardian.co.uk/education/2012/jan/09/computer-studies-in-schools

I did a little digging and found the following document on the General Teaching Council of England website – Annual digest of statistics 2010–11 – Profiles of registered teachers in England [PDF] – that contains demographic stats, amongst others, for UK teachers. But no stats relating to subject areas of degree level qualifications held, which is presumably the data referred to in the tweet. So I’m thinking: this is partly where the role of data journalist comes in… They may not be able to verify the numbers by checking independent sources, but they may be able to shed some light on where the numbers came from and how they were arrived at, and maybe even secure their release (albeit as a single point source?)