Thursday, 12 September 2019

Bombast meets dotty

The Beloved spends some of her time in pursuit of The Miracle of Mindfulness [it's a book]. It is mildly interesting, but not really surprising, that others in the same practice are angry, vindictive, spiteful and lacking in compassion [not all of them and not all the time you may be sure]. You'd expect better of them, of course but then again if you're a bit out of kilter [unbalanced would put it too strong] maybe that's why you [need to] come within the orbit of walking meditation and calming the mind. You'd expect that scientists would be rational, objective, fair, data-driven and able to change their minds when presented with new evidence. That's the theory, but ambition, competitive behaviour, selfishness, jealousy and prejudice are no less common in the scientific community than among tinkers, tailors [R] or sailors.

In 2013, I lost the rag at a Sunday Times report that ranked British and Irish Universities. The Institute continued for several years to puff itself up on the Best Institute of the Year 2014 despite there being objective evidence that other Irish Institutes were, by the SunTim criteria of judgement, objectively better. There are few things sadder than places which still sport plaques like Tidiest Town in the World 1967, after all the folk that made it so tidy are dead. My exec summ from the 2013 Ranking University exercise was "I want TCD (my Alma Mater) to win and my own employer did win, but it's still absolute bloody nonsense".

That was then, this is now. John Ioannidis has assembled a crack team to rank researchers by their success in the business of publishing their work. There are 6 million published, publishing scientists across the world. There is little point in doing what they are doing, let alone get paid for it, unless other researchers read the stuff, incorporate it into their own practice and cite the original study. I've had occasion to look at the science of citation before and before that. I've also had a slap at the academic publishers who lubricate, and profit from, the process.  The paper itself is only 6 pages long but they have made available citation data on thousands of named scientists. The basic Excel spreadsheet, with 40 columns of metrics, is 34MB is size; which would take 3 days to download on our wireless 'broadband' connection. Accordingly, I downloaded the file at work, used the sort function and clipped out just the Irish data, which is small enough to send as an e-mail attachment. I love data.

I probably wouldn't have heard about this study, if my ear to the ground in Trinity College Dublin hadn't sent me a fragment of tendentious 'analysis' of these data which appeared to show that TCD was a) wonderful but b) could do better. At the very least, the management of TCD should pay attention to the study and do something about it. When I 'analysed' the Sunday Times data in 2013 I found that they had failed to sum columns of figures correctly, and, by their own criteria, given the Best Institute gong to the wrong place. If there are obvious, checkable, correctable errors in data then the safest response is to ignore the whole thing because Quality Control has clearly failed.

The Ioannidis data is available for public download and discourse. Indeed people are encouraged to get to grips with the numbers and do some critical analysis. As I say, I'm limiting my investigations to Irish scientists because, frankly Scarlett, I know nothing about the research stars of Taiwan or Afghanistan. And <brrrzzzzz>, the top cited researcher for Ireland, Luke O'Neill, on the first row most people will look at, is wrongly housed in the Technological University of Dublin TUD rather than TCD! Sorting the data revealed several cases of duplicate entries: Coey J.M.D. (another star TCD player) has a doppleganger called Coey JMD, for the most obvious example. This is why I think we should be super skeptical about data if/before drawing any conclusions from it. If 'they' can't get the names or institutions sorted [easy] how much confidence would you have in the fact that, for O'Neill "hm18 (ns) = 52.79329482". Glaaaarrrrkkk. What kind of table reports woolly metrics like the HM-index to 8 significant figures? I knock that out our students by second year.  Blaa blaa blaa.

I worked for 14 years in TCD and 7 years in UCD, 3 of which I was double jobbing in both places at the same time. So I'm interested in the longtime rivalry between the two most internationally profiled Universities in Ireland. One of the TCD analysts was super chuffed that the top 3 most cited researchers for 2018 were all from TCD <puff puff preen preen> but when you look at the data from 2017 TCD secures places 1, 4, 5 because UL's Zaworotko, Michael and UCC's Shanahan, Fergus bounced them off the podium. Those West Country interlopers are 4 and 5 in 2018. It just shows that these metrics are fluid and a bit chaotic and unreliable. No university can depend on a single genius for its reputation unless it be Professor Harry Potter, Master of Quiddich and the Universe.  If you just look at one year and score yourself as 1 2 3 and think that sweeps the board then you are guilty of selection bias - stopping when the result is most favourable for your hypothesis. That is a most egregious scientific sin. Why stop at 3?? If you look at the 10 most cited people from TCD (for 2018 to give TCD its best shot); then they rank as
TCD 1 2 3 11 17 20 21 24 31 32 sum = 162
compared to ranks for the top ten for
UCD 6 7 8 9 12 13 15 26 29 30 sum = 155
smaller is better and this is better than TCD
You may bet that the Publicity suits in UCD will talk about The Top Ten whereas TCD with prefer a gold, silver and bronze metaphor.  But they are both spouting nonsense because the difference in cumulative scores is not statistically significant. A dot-plot [again!] is really useful in showing this:
That illustrates the World ranks for the best workers at the two institutions. The best person the Irish education /research system is  # 477 on the world stage and Ioannidis drew a line so that lowest Dublin scientist recorded [at #129749) is John T Sheridan who studies optics at UCD. Me, I am nowhere! Not even among the grass under the feet of these giants . . . a speck, a microbe. John Sheridan is a really successful productive scientist and he almost didn't make the cut. But here's a peculiar personal thing - Joe Duffy is UCD's Most Cited Researcher in these lists; in 2003 I co-authored a paper with Joe which has accumulated 93 citations.  In a small-small sense, I have made a contribution to Ireland's success in these polls.

But it's not about me. I sent my dot-plot to a bunch of friends (incl some more co-authors) from TCD and UCD with the message "The research outputs of your two prestigious institutes are identical, you really should choke down your sibling rivalry and get married".  Such a merger has been the elephant in the room for at least 50 years.

No comments:

Post a Comment