I've been playing with the #REF2021 data - the massive, funding-defining, UK research assessment exercise which provides a (faulty-metric defined) glimpse of structure, size and inequality in UK University research system.

The data is open results2021.ref.ac.uk To visualise each institution's research make a square for each discipline/department they submitted ("UoAs"), scaled by the staff in that UoA, and coloured by the GPA that unit received for their research ("hot" colourmap - more white/yellow is numbers closer to 4, the maximum)

Do you have from a favourite cup or mug? And do you drink caffeinated beverages? Help me check something please: will need to RT this for the evening crowd later, since I presume answers are currently suffering from major morning-person sample bias

(my theory currently holding up better than expected, though)

I have just found some cognitive psychology lecture notes from 1943 in a hidden room of the old departmental annex. They show what @sheffielduni psychology students were learning nearly exactly 78 years ago @AnnaKClements you will appreciate why this has made me slightly late for our meeting!

Here's the (much delayed) showcase of student submissions to my module on data management and visualisation. Class of 2021 : tomstafford.github.io/psy6422/class-…

(All shared with permission, the module is designed so students create a public data project for a portfolio item) Most students start this course without any coding experience (and a few of them dread coding, as you'd expect). Almost universally they end proud of what they've learnt and surprising themselves with what they've done.

Observing Many Researchers using the Same Data and Hypothesis Reveals a Hidden Universe of Data Analysis osf.io/preprints/meta… Preprint
- 162 researchers (73 teams)
- data: six questions from the International Social Survey Programme
- hypothesis: whether immigration reduces public support for social policies
- results: massive variation, not predicted by researcher skills or beliefs

Today: attending (and talking at) this

https://twitter.com/SheffMetaNet/status/1367051143289962499

My slides are here tomstafford.staff.shef.ac.uk/talks/ the highlight is the slide which combines reference to statistical theory, Jorge Luis Borges and a certain classic Jim Henson movie from the 1980s

Academic colleagues, @LivUni are targeting 47 colleagues with redundancy, after an abrupt, non-consultative and opaque review which uses weaponised grant income and citation metrics. Please sign and share this open letter to protest
docs.google.com/document/d/1OJ… #LiverpoolRedundancies Form for signing is here (you can't edit the letter directly)
docs.google.com/forms/d/e/1FAI… #LiverpoolRedundancies

1. Bitcoin is destroying the planet
2. Reviewing papers is unrecognised and unrewarded
So (and hear me out here)
3. We invent a new cryptocurrency which you 'mine' by contributing quality reviews of the literature APCs for journal articles are replaced with KudosCoin fees, which you can either generate by reviewing or buy off people who do

Today: listening to this

https://twitter.com/G_Karadzhov/status/1361338176045019138

Love this analysis of vulnerabilities to Wikipedia's knowledge integrity from this paper: arxiv.org/abs/1910.12596

I am reading Judea Pearl's "Book of Why" on why statistical models need a causal framework to guide their interpretation. The explanatory examples he gives (p157ff) are twisting my noodle Fortunately I could use @LisaDeBruine's faux package to simulate some data which helped debruine.github.io/faux/index.html it was all delightfully easy and only took me about 10 minutes of hacking around in R to convince myself of what Pearl meant (his "Game 1" since you ask)

Interested in the phenomenology of why some students find turning their video on in seminars so aversive. Any thoughts? If this is you, can you describe why to me in a way I'll get? Is it related to the way some (mostly younger?) people find answering phone calls unbearable?

Students told to analyse a dataset to test a specific hypothesis tended to miss the gorilla hidden in the data genomebiology.biomedcentral.com/articles/10.11… students not given a hypothesis to test, just told to examine the data more likely to find the gorilla. Fiendish.

Pre-print with more details here biorxiv.org/content/10.110…

Why would you be driven to make baseless claims about election fraud? it seems so norm-eroding, so dangerous a precedent to set - knocking out the epistemic ladder up which we climb to democracy, leaving a future in which legitimacy can't exist. Presumably (1) you would only do so if you felt that the threat posed by losing the election was as large as the threat of the collapse of legitimate government anyway (i.e. if you were really desperate) but also ...

Who reads preprints? Academics and...er... white nationalists?? journals.plos.org/plosbiology/ar… "a framework for segmenting a scholarly article’s audience on Twitter...into granular, informative categories inferred through probabilistic topic modeling of metadata collected from each user’s network of followers"
(could be useful @nikaletras ?)

During my PhD I nearly lost my mind thinking about why we build computational models in cognitive science. {thread} Modelling defines what makes cognitive science different from psychology. All cognitive scientists know that formal, computable models are good, but we don't always say exactly *why*

And when we do say, we find we don't exactly agree

Legit learning outcomes for my new MSc module on data management and visualisation, right? Another module development question: should I devote an entire 2 hour class to how to google Stack Overflow, or rush it and cover it in 1 hour?

Thread, top five drone takedowns.
#5 Shotgun - simple, effective. A classic #4 Eagle - Dutch authorities *were* training eagles to take down drones, but have since given up on this idea (either due to risk of injury, or difficulty training the eagles, reports vary) theverge.com/2017/12/12/167…

Many Analysts, One Data Set: Making Transparent How Variations in Analytic Choices Affect Results journals.sagepub.com/doi/10.1177/25… our 65 author collaboration is finally published in Advances in Methods and Practices in Psychological Science As featured in this from @FiveThirtyEight "Science Isn’t Broken :It’s just a hell of a lot harder than we give it credit for." fivethirtyeight.com/features/scien…

Facebook research: despite improvements in statistical techniques for causal inference from observational data, and even with massive statistical power, observational methods fail to capture causal effects revealed by direct experiment kellogg.northwestern.edu/faculty/gordon… "Using data from 15 US advertising experiments at Facebook comprising 500 million user-experiment observations and 1.6 billion ad impressions..."