dan crouch (@djmcrouch) profile picture

Genetics/Statistics/Evolution researcher

UK
Joined July 2017
Photo by dan crouch (@djmcrouch)Photo by dan crouch (@djmcrouch)Photo by dan crouch (@djmcrouch)Photo by dan crouch (@djmcrouch)Photo by dan crouch (@djmcrouch)Photo by dan crouch (@djmcrouch)Photo by dan crouch (@djmcrouch)Photo by dan crouch (@djmcrouch)Photo by dan crouch (@djmcrouch)
In this short article I propose an alternative way to use P-values, for quantifying the probability of making an false-positive error in the future if one were propose that the alternative hypothesis was true based on the observed P-value pnas.org/doi/10.1073/pnas.24…
1
2
259
Interview with Yoav Benjamini. At around 5 minutes in he tells the story of how the False-Discovery Rate was published. It took 5 years of submissions and reviews, across three different journals, before it was accepted! m.youtube.com/watch?v=ZGU4Sl…
71
The approach is to compute the chance of incorrectly finding further evidence for the alternative hypothesis when the null hypothesis is true, so I call it the 'False Evidence Rate'
1
61
Unlike traditional P-value interpretations, it provides an error rate that depends on the precise value of the observed P-value, but also does not require any Bayesian assumptions
49
Something that happened to me a few times: a) design stats method, b) test on sims, c) worry for a long time when you notice it not performing perfectly, d) run 10x more simulations and realise the errors were noise Trust in frequentism: it works in the long run!
1
103
Whatever your historically based political sympathies, if you have even a shred of human decency you surely cannot morally justify Saturday's barbaric attacks on innocent civilians. “A plague o’ both your houses” is, for once, inappropriate.
865
1,777
233
15,630
1,999,961
Is having even more gene-disease associations still useful for picking drug targets? For which types of programs does it matter? Has pharma shifted focus towards genetically validated targets? New paper by me, @mnelsonxy, @DongCoco90417, & @jivecast doi.org/10.1101/2023.06.23.2…👇
4
79
10
245
169,396
Do you work in biomedical #research? Do you also question your #antibodies? Dodgy antibodies waste time & $$$ + drive the reproducibility crisis. Let's talk about it! We've validated >50❗️Tau antibodies: rb.gy/jf247 Here's what we learnt & why it matters A 🧵 1/n
6
46
14
140
36,262
On proper reflection this looks like an interesting and useful study, but my initial reaction was 'no shit'
'Disruptive' science has plummeted — and no one knows why. Research in the 2000s is much more likely to incrementally push science forward than to veer off in a new direction. Check out my latest for @Nature with @russellfunk, @dashunwang, @yian_yin nature.com/articles/d41586-0…
1
1
315
Replying to @VFSGlobal
@VFSGlobal We are having difficulty arranging two visa appointments at the Cebu UK visa centre. Whenever I select a free appointment time and progress to payment I get the message...
2
208
...'An error has encountered, please try again after sometime. Alternatively, please check the Dashboard for booked appointments before retrying'. Any chance you can send me a customer services contact email? Thanks
109
Nice piece of work on common issues with interpretations of Bayes factors found in papers. psyarxiv.com/du3fc/
1
I think some of these 'questionable practices' are possibly forgivable in the sense that they are smallish inaccuracies servicing easier communication but practices that involve improper reporting of the prior create big issues
Among the measures I could think of to improve academia in this respect, would be an upper cap for the number of publications every individual scientist can co-author each year (e.g. 24). Incidentally, this would likely improve the average quality of the scientific literature. 4/
20
11
3
103
Check out my API for comparing two sets of ratings associated with online reviews. It estimates the posterior probability that one item is truly better-rated, on average, than the other, using a model fit to a large database of Amazon ratings. rapidapi.com/djmcrouch/api/r…
2
7
Jurassic world dominion is full of pseudoscience, bad dialogue and poorly lit, ugly action sequences. Big disappointment, and I even like Jurassic Park 3!
1
dan crouch retweeted
The recordings from #EMGM2022 are now ready, organised by session. Check them out here: youtube.com/playlist?list=PL…
10
18
The skeptic community is fond of the saying 'extraordinay claims require extraordinary evidence'. The problem with this is we have an agreed-upon system for determining what is extraodinary evidence (statistics), but who decides what is an extraordinary claim?
1
2