"It seems we're undersampling a metric shit-ton of people who live in Michigan"
November 4, 2024 at 12:55 AM
"It seems we're undersampling a metric shit-ton of people who live in Michigan"
it turns out UMich consumer sentiment being so negative may have been because of a polling method swap + undersampling of Republicans
www.natesilver.net/p/is-the-vib...
www.natesilver.net/p/is-the-vib...
Is the vibecession real — or is the survey broken?
A shift to online polling and undersampling of Republicans are skewing America’s most-cited measure of consumer sentiment.
www.natesilver.net
July 2, 2026 at 7:21 PM
it turns out UMich consumer sentiment being so negative may have been because of a polling method swap + undersampling of Republicans
www.natesilver.net/p/is-the-vib...
www.natesilver.net/p/is-the-vib...
Undersampling women and Black people in MI from a Michigan-based poll is... it's something.
November 4, 2024 at 12:46 AM
Undersampling women and Black people in MI from a Michigan-based poll is... it's something.
Yeah to be clear I think this poll is terrible news for the GOP nationally it's just that I would bet everything they're undersampling rural Arkansas
August 19, 2026 at 11:56 PM
Yeah to be clear I think this poll is terrible news for the GOP nationally it's just that I would bet everything they're undersampling rural Arkansas
he’s seeing something in his results and probably comparing with the voter file that tells him he’s undersampling them i assume
November 4, 2024 at 12:58 AM
he’s seeing something in his results and probably comparing with the voter file that tells him he’s undersampling them i assume
If pollsters would simply offer some kind of scam when they called people, their problems with undersampling Trump voters would be over.
February 12, 2025 at 2:16 PM
If pollsters would simply offer some kind of scam when they called people, their problems with undersampling Trump voters would be over.
New preprint from our group! We propose SHAPES, a set of metrics to quantify the distributional coverage of generative models of protein structures with embeddings at different structural hierarchies and quantify undersampling / extrapolation behaviors.
January 15, 2025 at 6:48 PM
New preprint from our group! We propose SHAPES, a set of metrics to quantify the distributional coverage of generative models of protein structures with embeddings at different structural hierarchies and quantify undersampling / extrapolation behaviors.
The latest thin reed is trumpeting/bemoaning (depending on their political persuasion) that polling must be systemically oversampling progressives & undersampling the mythical Shy Trump Voter™️bc polling in WI didnt capture v late movement—as if we dont have election results saying "Rs are fucked"
August 18, 2026 at 2:30 PM
The latest thin reed is trumpeting/bemoaning (depending on their political persuasion) that polling must be systemically oversampling progressives & undersampling the mythical Shy Trump Voter™️bc polling in WI didnt capture v late movement—as if we dont have election results saying "Rs are fucked"
CODA mapping reveals that 100% of fallopian tubes of healthy women contain precancers!
The standard SEE-FIM protocol misses most of them due to undersampling. But our exhaustive assessment reveals that both pre- and postmenopausal women have them.
Here: www.biorxiv.org/content/10.1...
The standard SEE-FIM protocol misses most of them due to undersampling. But our exhaustive assessment reveals that both pre- and postmenopausal women have them.
Here: www.biorxiv.org/content/10.1...
April 25, 2026 at 7:11 PM
CODA mapping reveals that 100% of fallopian tubes of healthy women contain precancers!
The standard SEE-FIM protocol misses most of them due to undersampling. But our exhaustive assessment reveals that both pre- and postmenopausal women have them.
Here: www.biorxiv.org/content/10.1...
The standard SEE-FIM protocol misses most of them due to undersampling. But our exhaustive assessment reveals that both pre- and postmenopausal women have them.
Here: www.biorxiv.org/content/10.1...
If you want to map the spatial composition of a tissue or organ, 3D mapping is not an option.
It is absolutely necessary.
Due to its inherent undersampling, 2D omic mapping is almost always misleading.
See details here: www.sciencedirect.com/science/arti...
It is absolutely necessary.
Due to its inherent undersampling, 2D omic mapping is almost always misleading.
See details here: www.sciencedirect.com/science/arti...
December 24, 2025 at 10:32 AM
If you want to map the spatial composition of a tissue or organ, 3D mapping is not an option.
It is absolutely necessary.
Due to its inherent undersampling, 2D omic mapping is almost always misleading.
See details here: www.sciencedirect.com/science/arti...
It is absolutely necessary.
Due to its inherent undersampling, 2D omic mapping is almost always misleading.
See details here: www.sciencedirect.com/science/arti...
"We're undersampling more than half the state's population" seems like a weird thing to admit? Like I'm not a pollster but that's how the below comes off to me.
November 4, 2024 at 1:23 AM
"We're undersampling more than half the state's population" seems like a weird thing to admit? Like I'm not a pollster but that's how the below comes off to me.
This poll is oversampling homeowners by about 13 points and undersampling renters by around 15 points.
October 27, 2025 at 4:27 PM
This poll is oversampling homeowners by about 13 points and undersampling renters by around 15 points.
I think the unusually long branches in the BA.3.2 tree indicate two things:
#1. Slow growth globally—fast growth results in many identical sequences, if surveillance is sufficient
#2. Undersampling—BA.3.2 likely most common in poorer regions of world where there's little sequencing of late. 3/3
#1. Slow growth globally—fast growth results in many identical sequences, if surveillance is sufficient
#2. Undersampling—BA.3.2 likely most common in poorer regions of world where there's little sequencing of late. 3/3
July 2, 2025 at 11:59 AM
I think the unusually long branches in the BA.3.2 tree indicate two things:
#1. Slow growth globally—fast growth results in many identical sequences, if surveillance is sufficient
#2. Undersampling—BA.3.2 likely most common in poorer regions of world where there's little sequencing of late. 3/3
#1. Slow growth globally—fast growth results in many identical sequences, if surveillance is sufficient
#2. Undersampling—BA.3.2 likely most common in poorer regions of world where there's little sequencing of late. 3/3
I think all mention of the Lord Ashcroft poll should note the glaring undersampling of the rural (less Reform friendly) third of the constituency. It accounts for just 2% of the weighted sample!
New post just out:
My preview of the local elections which includes predictions for every council, mayoral election, and the Runcorn by-election.
I've sub-titled it "Tory cataclysm edition" which will give you a bit of a clue...
(£/free trial)
open.substack.com/pub/samf/p/l...
My preview of the local elections which includes predictions for every council, mayoral election, and the Runcorn by-election.
I've sub-titled it "Tory cataclysm edition" which will give you a bit of a clue...
(£/free trial)
open.substack.com/pub/samf/p/l...
Local Elections 2025: The Preview
Tory cataclysm edition
open.substack.com
April 23, 2025 at 10:31 AM
I think all mention of the Lord Ashcroft poll should note the glaring undersampling of the rural (less Reform friendly) third of the constituency. It accounts for just 2% of the weighted sample!
Again, ultra-relevant paper to the #eDNA crowd - I find *so many* marine species just outside their known/species distribution model range, because of undersampling in the literature plus ongoing poorly understood species range shifts due to climate change
Excited to announce our new paper, “Global impoverishment of natural vegetation revealed by dark diversity,” out today in Nature – and it’s open access! Huge global collaboration led by Meelis Pärtel 🌐@macroecologyut.bsky.social
www.nature.com/articles/s41...
www.nature.com/articles/s41...
Global impoverishment of natural vegetation revealed by dark diversity - Nature
A comparison of alpha diversity (number of plant species) and dark diversity (species that are currently absent from a site despite being ecologically suitable) demonstrates the negative effects of re...
www.nature.com
April 7, 2025 at 5:17 AM
Again, ultra-relevant paper to the #eDNA crowd - I find *so many* marine species just outside their known/species distribution model range, because of undersampling in the literature plus ongoing poorly understood species range shifts due to climate change
I have a new post up at Nate Silver's site about UMich's sentiment survey. Please read it I think it's quite good. The upshot is that the survey is broken because it doesn't weight by party. If you're talking vibecession, get a different survey. www.natesilver.net/p/is-the-vib...
Is the vibecession real — or is the survey broken?
A shift to online polling and undersampling of Republicans are skewing America’s most-cited measure of consumer sentiment.
www.natesilver.net
June 29, 2026 at 5:41 PM
I have a new post up at Nate Silver's site about UMich's sentiment survey. Please read it I think it's quite good. The upshot is that the survey is broken because it doesn't weight by party. If you're talking vibecession, get a different survey. www.natesilver.net/p/is-the-vib...
Yeah I’m continuing to believe (probably cope) that the polls are wildly undersampling young people
August 13, 2024 at 8:14 PM
Yeah I’m continuing to believe (probably cope) that the polls are wildly undersampling young people
One thing that happened over the past two years is the survey broke and this persistent downward trend is fake. www.natesilver.net/p/is-the-vib...
July 23, 2026 at 11:45 PM
One thing that happened over the past two years is the survey broke and this persistent downward trend is fake. www.natesilver.net/p/is-the-vib...
🚨vConTACT3 preprint live!🚨(Peer Review soon...!)
vConTACT3 delivers a unified, scalable, and transparent framework for genome-based virus taxonomy — helping translate big viral data into systematic classification.
🔗 Read the preprint: doi.org/10.1101/2025...
Improvements details below 👇
vConTACT3 delivers a unified, scalable, and transparent framework for genome-based virus taxonomy — helping translate big viral data into systematic classification.
🔗 Read the preprint: doi.org/10.1101/2025...
Improvements details below 👇
Scalable and systematic hierarchical virus taxonomy with vConTACT3
Viruses are key players in diverse ecosystems, but studying their impacts is technically and taxonomically challenging. Taxonomic complexities derive from undersampling, diverse DNA and RNA genomes wi...
doi.org
November 7, 2025 at 4:36 PM
🚨vConTACT3 preprint live!🚨(Peer Review soon...!)
vConTACT3 delivers a unified, scalable, and transparent framework for genome-based virus taxonomy — helping translate big viral data into systematic classification.
🔗 Read the preprint: doi.org/10.1101/2025...
Improvements details below 👇
vConTACT3 delivers a unified, scalable, and transparent framework for genome-based virus taxonomy — helping translate big viral data into systematic classification.
🔗 Read the preprint: doi.org/10.1101/2025...
Improvements details below 👇
Thanks this really points to undersampling. Luckily things get better overtime with potential training data.
November 27, 2025 at 12:38 PM
Thanks this really points to undersampling. Luckily things get better overtime with potential training data.
A lot of R and cooked polls do this too, I think they don't want to be accused of undersampling Ds (?) so they steal independents for more Rs
Yeahhhh R internal bias showing right there
Yeahhhh R internal bias showing right there
September 11, 2026 at 9:47 PM
A lot of R and cooked polls do this too, I think they don't want to be accused of undersampling Ds (?) so they steal independents for more Rs
Yeahhhh R internal bias showing right there
Yeahhhh R internal bias showing right there
We call that "inadequate biodiversity baselines preclude the accurate assessment of Areas of Particular Environmental Interest due to undersampling of large mobile fauna"
February 14, 2026 at 10:36 PM
We call that "inadequate biodiversity baselines preclude the accurate assessment of Areas of Particular Environmental Interest due to undersampling of large mobile fauna"
We do better than 0.5 au with a camera not designed for the this work (4 arsec pixels and severe undersampling). We can easily do 10x better with cameras being contemplated. But I don't know the real answer. Someone else might know, though.
June 30, 2025 at 4:41 AM
We do better than 0.5 au with a camera not designed for the this work (4 arsec pixels and severe undersampling). We can easily do 10x better with cameras being contemplated. But I don't know the real answer. Someone else might know, though.