Dylan Williams
@dylwil.bsky.social
Alzheimer's Research UK Senior Fellow, University College London. Studying the molecular epidemiology of dementia.
Nice piece, but you miss the most important insight....that we shouldn't be doing heritability analysis at all. They are pointless and highly misleading... academic.oup.com/ije/article-... pmc.ncbi.nlm.nih.gov/articles/PMC...
Epidemiology in a changing world: variation, causation and ubiquitous risk factors
Abstract. We are all living in the era of globalization and, like it or not, it is going to change the way we practise epidemiology, the kinds of questions
academic.oup.com
November 25, 2025 at 11:59 AM
Nice piece, but you miss the most important insight....that we shouldn't be doing heritability analysis at all. They are pointless and highly misleading... academic.oup.com/ije/article-... pmc.ncbi.nlm.nih.gov/articles/PMC...
You do have to wonder if @mendelrandom.bsky.social has gone round leaving copies in every pub in Bristol though...
November 3, 2025 at 3:56 PM
You do have to wonder if @mendelrandom.bsky.social has gone round leaving copies in every pub in Bristol though...
Alas, my data vis is not that pretty
August 12, 2025 at 1:46 PM
Alas, my data vis is not that pretty
Which is when by the way? I will come to Bristol for it if I haven't missed it!
August 12, 2025 at 1:44 PM
Which is when by the way? I will come to Bristol for it if I haven't missed it!
That could be the title of your inaugural lecture?
August 12, 2025 at 1:44 PM
That could be the title of your inaugural lecture?
@michelnivard.bsky.social you're at least a bluefin tuna of statistical genetics mate!
August 12, 2025 at 1:30 PM
@michelnivard.bsky.social you're at least a bluefin tuna of statistical genetics mate!
@erictopol.bsky.social That is quite a strong conclusion to make from a subgroup analysis with N=29 from a phase 2 trial (and where, IMO, such large effect size stretches plausibility)
August 1, 2025 at 3:13 PM
@erictopol.bsky.social That is quite a strong conclusion to make from a subgroup analysis with N=29 from a phase 2 trial (and where, IMO, such large effect size stretches plausibility)
Arising for a few reasons inc little incidence in 'unexposed' group (e2 homozygotes), limited outcome ascertainment in the cohort, imprecision for PAFs at each exposure level aggregating. Less problematic considering large point estimates for the PAFs produced every time (and a strong prior here)
May 27, 2025 at 8:27 AM
Arising for a few reasons inc little incidence in 'unexposed' group (e2 homozygotes), limited outcome ascertainment in the cohort, imprecision for PAFs at each exposure level aggregating. Less problematic considering large point estimates for the PAFs produced every time (and a strong prior here)
...Some feedback from last year’s course:
“The entire course was excellent. Lots of helpful information, engaging content, very helpful and detailed practicals.”
“The course leaders were all super helpful, very kind and very knowledgeable - it was a great atmosphere.”
“The entire course was excellent. Lots of helpful information, engaging content, very helpful and detailed practicals.”
“The course leaders were all super helpful, very kind and very knowledgeable - it was a great atmosphere.”
April 28, 2025 at 9:34 AM
...Some feedback from last year’s course:
“The entire course was excellent. Lots of helpful information, engaging content, very helpful and detailed practicals.”
“The course leaders were all super helpful, very kind and very knowledgeable - it was a great atmosphere.”
“The entire course was excellent. Lots of helpful information, engaging content, very helpful and detailed practicals.”
“The course leaders were all super helpful, very kind and very knowledgeable - it was a great atmosphere.”
..This covers one-sample and two-sample approaches, sensitivity analyses including pleiotropy-robust methods, multivariable MR for mediation and interaction, within-family MR, MR-PheWAS and drug target MR...
April 28, 2025 at 9:34 AM
..This covers one-sample and two-sample approaches, sensitivity analyses including pleiotropy-robust methods, multivariable MR for mediation and interaction, within-family MR, MR-PheWAS and drug target MR...
...many component causes of AD either directly interact or converge on the same mechanisms. So attributable fractions for them overlap and adding up combinations of can quickly sum to more than 100%. I.e. AD causes are not mutually exclusive slices of risk
April 25, 2025 at 1:41 PM
...many component causes of AD either directly interact or converge on the same mechanisms. So attributable fractions for them overlap and adding up combinations of can quickly sum to more than 100%. I.e. AD causes are not mutually exclusive slices of risk