the polls were right, and undersampling Democrats, if anything
November 4, 2026 at 3:02 PM
the polls were right, and undersampling Democrats, if anything
Idk the fact that the ME-02 numbers are good for Republicans makes me think this poll isn't undersampling rural voters
November 2, 2026 at 2:48 AM
Idk the fact that the ME-02 numbers are good for Republicans makes me think this poll isn't undersampling rural voters
i think the polls are significantly undersampling AES voters and it might be a large undersampling, because um the guy got 500 people in Keweenaw County, a county of about 2000 people and nobody in that audience read stuffer.
the ratfucking is gonna fail, perhaps massively. can't wait!
the ratfucking is gonna fail, perhaps massively. can't wait!
October 31, 2026 at 9:25 AM
i think the polls are significantly undersampling AES voters and it might be a large undersampling, because um the guy got 500 people in Keweenaw County, a county of about 2000 people and nobody in that audience read stuffer.
the ratfucking is gonna fail, perhaps massively. can't wait!
the ratfucking is gonna fail, perhaps massively. can't wait!
StrokeGuard: Cloud-Deployed Calibrated-MLP with K-Means Boundary Undersampling for Scalable Stroke Risk Prediction #NeuroDegeneration 🧪🧠
https://www.researchsquare.com/article/rs-11286884/latest
https://www.researchsquare.com/article/rs-11286884/latest
November 2, 2026 at 8:02 AM
StrokeGuard: Cloud-Deployed Calibrated-MLP with K-Means Boundary Undersampling for Scalable Stroke Risk Prediction #NeuroDegeneration 🧪🧠
https://www.researchsquare.com/article/rs-11286884/latest
https://www.researchsquare.com/article/rs-11286884/latest
yeah, the not blue enough scenario does sound too good to be true imo.
I'd buy some R undersampling. idk if I buy that it's as much as in a Trumpy year though, if so.
we shall see.
I'd buy some R undersampling. idk if I buy that it's as much as in a Trumpy year though, if so.
we shall see.
November 2, 2026 at 4:37 AM
yeah, the not blue enough scenario does sound too good to be true imo.
I'd buy some R undersampling. idk if I buy that it's as much as in a Trumpy year though, if so.
we shall see.
I'd buy some R undersampling. idk if I buy that it's as much as in a Trumpy year though, if so.
we shall see.
Is the water surface really sub-micron flatness? Or is there a different relevant length scale due to the undersampling?
October 29, 2026 at 10:01 PM
Is the water surface really sub-micron flatness? Or is there a different relevant length scale due to the undersampling?
"Oh jinkies how do you know that's what they meant by 'mail/already voted'?"
GIRL I CALLED THEM
I CALLED THEM FOR THIS REASON. SEE A PARAPHRASE OF HOW THE CALL WENT HERE.
IT IS CERTAIN THEY ARE UNDERSAMPLING MAIL. BUT IT DOESNT MATTEr.
bsky.app/profile/jinx...
GIRL I CALLED THEM
I CALLED THEM FOR THIS REASON. SEE A PARAPHRASE OF HOW THE CALL WENT HERE.
IT IS CERTAIN THEY ARE UNDERSAMPLING MAIL. BUT IT DOESNT MATTEr.
bsky.app/profile/jinx...
Basically:
I call literally at 6:09 at night expecting to leave a message, hear nothing back ever.
Gruff old dude picks. "Hello?" he asks.
"Hi I'm a Michigan citizen who likes to do amateur data modeling" I reply "And I have a question about a poll"
I call literally at 6:09 at night expecting to leave a message, hear nothing back ever.
Gruff old dude picks. "Hello?" he asks.
"Hi I'm a Michigan citizen who likes to do amateur data modeling" I reply "And I have a question about a poll"
October 28, 2026 at 11:03 PM
"Oh jinkies how do you know that's what they meant by 'mail/already voted'?"
GIRL I CALLED THEM
I CALLED THEM FOR THIS REASON. SEE A PARAPHRASE OF HOW THE CALL WENT HERE.
IT IS CERTAIN THEY ARE UNDERSAMPLING MAIL. BUT IT DOESNT MATTEr.
bsky.app/profile/jinx...
GIRL I CALLED THEM
I CALLED THEM FOR THIS REASON. SEE A PARAPHRASE OF HOW THE CALL WENT HERE.
IT IS CERTAIN THEY ARE UNDERSAMPLING MAIL. BUT IT DOESNT MATTEr.
bsky.app/profile/jinx...
Yeah this one has this
It's sad from SN though, I really trusted them, Chaz is letting his GOP undersampling instincts take the lead it would seem. He was right to ignore them last year
I'd be so curious how a Bluesky-only pollster would feel about these things
bsky.app/profile/hyde...
It's sad from SN though, I really trusted them, Chaz is letting his GOP undersampling instincts take the lead it would seem. He was right to ignore them last year
I'd be so curious how a Bluesky-only pollster would feel about these things
bsky.app/profile/hyde...
High school education or less voters appear to be 10% more of the electorate than in 2022, a bit strange to predict. Higher 65+ than in 2022 also sticks out some
I think all this at the margins makes this fine for Hinojosa still, but I may be off on her margin if this is closer than YouGov Blue
I think all this at the margins makes this fine for Hinojosa still, but I may be off on her margin if this is closer than YouGov Blue
October 29, 2026 at 12:47 PM
Yeah this one has this
It's sad from SN though, I really trusted them, Chaz is letting his GOP undersampling instincts take the lead it would seem. He was right to ignore them last year
I'd be so curious how a Bluesky-only pollster would feel about these things
bsky.app/profile/hyde...
It's sad from SN though, I really trusted them, Chaz is letting his GOP undersampling instincts take the lead it would seem. He was right to ignore them last year
I'd be so curious how a Bluesky-only pollster would feel about these things
bsky.app/profile/hyde...
even with the very close polls, I was thinking they were certainly undersampling Native Americans. Granted, that's only a small percentage of the vote, but as we've well seen, a small percentage can be the difference.
October 28, 2026 at 1:09 PM
even with the very close polls, I was thinking they were certainly undersampling Native Americans. Granted, that's only a small percentage of the vote, but as we've well seen, a small percentage can be the difference.
also a reflexive tendency to worry about polls undersampling gop voters, which is understandable given past experiences(!) but really doesn't seem supported; if anything, it's a coin flip between "samples are right-on, things are bad for GOP", and "republicans were a bit oversampled; it's a wipeout"
the idea Republican turnout will recover is just a baseline assumption. it sounds familar.
October 28, 2026 at 9:51 PM
also a reflexive tendency to worry about polls undersampling gop voters, which is understandable given past experiences(!) but really doesn't seem supported; if anything, it's a coin flip between "samples are right-on, things are bad for GOP", and "republicans were a bit oversampled; it's a wipeout"
Correcting class imbalance can make your risk model worse.
In this study, undersampling, oversampling and SMOTE produced badly miscalibrated probabilities, without improving discrimination.
A must read if you build risk models 👇
In this study, undersampling, oversampling and SMOTE produced badly miscalibrated probabilities, without improving discrimination.
A must read if you build risk models 👇
The harm of class imbalance corrections for risk prediction models: illustration and simulation using logistic regression
Methods to correct class imbalance (imbalance between the frequency of outcome events and nonevents) are receiving increasing interest for developing prediction models. We examined the effect of imbalance correction on the performance of logistic ...
pmc.ncbi.nlm.nih.gov
October 29, 2026 at 8:02 AM
Correcting class imbalance can make your risk model worse.
In this study, undersampling, oversampling and SMOTE produced badly miscalibrated probabilities, without improving discrimination.
A must read if you build risk models 👇
In this study, undersampling, oversampling and SMOTE produced badly miscalibrated probabilities, without improving discrimination.
A must read if you build risk models 👇
Whatever you think of him personally, Silver has been very very good at forecasting elections.
October 27, 2026 at 4:36 PM
Whatever you think of him personally, Silver has been very very good at forecasting elections.
Oh I know, it baffles me why every US poll they do is just R+2
But undersampling conservatives seems to be an issue in Brazil more than it is here tbh
But undersampling conservatives seems to be an issue in Brazil more than it is here tbh
October 25, 2026 at 11:15 PM
Oh I know, it baffles me why every US poll they do is just R+2
But undersampling conservatives seems to be an issue in Brazil more than it is here tbh
But undersampling conservatives seems to be an issue in Brazil more than it is here tbh
Perhaps they're undersampling Republican support because they can't find any.
October 22, 2026 at 9:58 PM
Perhaps they're undersampling Republican support because they can't find any.
oh also, anyone worried about nonresponse bias by Republicans needs to read the State Navigate article, just scroll down for a bit
very very good method of determining that there is no undersampling of Republicans occurring. this is truly that more of them just don't want to vote / flipped
very very good method of determining that there is no undersampling of Republicans occurring. this is truly that more of them just don't want to vote / flipped
October 22, 2026 at 3:47 PM
oh also, anyone worried about nonresponse bias by Republicans needs to read the State Navigate article, just scroll down for a bit
very very good method of determining that there is no undersampling of Republicans occurring. this is truly that more of them just don't want to vote / flipped
very very good method of determining that there is no undersampling of Republicans occurring. this is truly that more of them just don't want to vote / flipped
Jim McLaughlin to Newsmax: Polls Are 'Undersampling Republicans'
Jim McLaughlin to Newsmax: Polls Are 'Undersampling Republicans'
Pollster Jim McLaughlin told Newsmax Wednesday that political surveys are deliberately "undersampling Republicans" in an effort to suppress GOP voter turnout and fundraising ahead of the midterm elections.
dlvr.it
October 22, 2026 at 1:31 AM
Jim McLaughlin to Newsmax: Polls Are 'Undersampling Republicans'
Johnson is flat wrong in asserting that the “polling organizations are being forced to admit that they are dramatically undersampling Republicans.” That’s just not true. One organization fabricated all of its data. That is not a “dramatic undersampling of Republicans.”
GOP panic over the midterms begins to damage their prospects for 2028
October 22, 2026
roberthubbell.substack.com
October 22, 2026 at 11:16 AM
Johnson is flat wrong in asserting that the “polling organizations are being forced to admit that they are dramatically undersampling Republicans.” That’s just not true. One organization fabricated all of its data. That is not a “dramatic undersampling of Republicans.”
Those polls are off when you check the crosstabs. The GA polls are undersampling Black voters by nearly 5 points. They based the numbers on 2024 turnout and so far Black voters are over-performing in 2026 turnout compared to 2024.
October 21, 2026 at 3:09 PM
Those polls are off when you check the crosstabs. The GA polls are undersampling Black voters by nearly 5 points. They based the numbers on 2024 turnout and so far Black voters are over-performing in 2026 turnout compared to 2024.
Sampling Rate: To cleanly sample a 1 MHz upper carrier frequency via undersampling or direct intermediate frequency (IF) sampling, you need a minimum of 2.5 to 5 MSPS per channel.
October 19, 2026 at 8:41 AM
Sampling Rate: To cleanly sample a 1 MHz upper carrier frequency via undersampling or direct intermediate frequency (IF) sampling, you need a minimum of 2.5 to 5 MSPS per channel.
Fascinating methodology, could be interesting to apply to other organisms but probably tough due to undersampling
Our latest work is now on bioRxiv! 🦠
ViroChart: Quantitative host-response mapping enables phenotypic profiling of viruses detected by virome sequencing.
www.biorxiv.org/content/10.6...
(1/9)
ViroChart: Quantitative host-response mapping enables phenotypic profiling of viruses detected by virome sequencing.
www.biorxiv.org/content/10.6...
(1/9)
Quantitative host-response mapping enables phenotypic profiling of viruses detected by virome sequencing
Animal-derived viruses have repeatedly caused human outbreaks, motivating efforts to identify and characterize viruses circulating in animals. Virome sequencing has accelerated animal virus discovery,...
www.biorxiv.org
October 18, 2026 at 8:14 PM
Fascinating methodology, could be interesting to apply to other organisms but probably tough due to undersampling
also TRAF of all people put out an R+3 poll.
In South Carolina.
With a trump +30 electorate and dramatically undersampling black voters.
Which kind of supports my theory since Lindsey was the biggest proponent of the Russian sanctions that she now can't use in her favor since it's now moot.
In South Carolina.
With a trump +30 electorate and dramatically undersampling black voters.
Which kind of supports my theory since Lindsey was the biggest proponent of the Russian sanctions that she now can't use in her favor since it's now moot.
October 19, 2026 at 3:45 AM
also TRAF of all people put out an R+3 poll.
In South Carolina.
With a trump +30 electorate and dramatically undersampling black voters.
Which kind of supports my theory since Lindsey was the biggest proponent of the Russian sanctions that she now can't use in her favor since it's now moot.
In South Carolina.
With a trump +30 electorate and dramatically undersampling black voters.
Which kind of supports my theory since Lindsey was the biggest proponent of the Russian sanctions that she now can't use in her favor since it's now moot.
Frank Luntz Just Admitted the Polls Still Can’t Find Trump Voters
https://zambianobserver.com/frank-luntz-just-admitted-the-polls-still-cant-find-trump-voters/
https://zambianobserver.com/frank-luntz-just-admitted-the-polls-still-cant-find-trump-voters/
Frank Luntz Just Admitted the Polls Still Can’t Find Trump Voters - The Zambian Observer
Frank Luntz Just Admitted the Polls Still Can’t Find Trump VotersCNN brought on longtime Republican pollster Frank Luntz and asked him the question every conservative has been asking for a decade: Are the polls undersampling the non-engaged Trump voters who don’t pick up the phone? Luntz didn’t dodge it. He owned the miss that made
zambianobserver.com
October 19, 2026 at 7:59 AM
Frank Luntz Just Admitted the Polls Still Can’t Find Trump Voters
https://zambianobserver.com/frank-luntz-just-admitted-the-polls-still-cant-find-trump-voters/
https://zambianobserver.com/frank-luntz-just-admitted-the-polls-still-cant-find-trump-voters/
This report is such an excellent breakdown of why polls were so off in 2025 bc modeling decisions undersampling youth voters & Dems, in spite of current data, engineered bad polls overestimating Republicans
And pollsters are continuing the same mistake now in 2026
And pollsters are continuing the same mistake now in 2026
October 16, 2026 at 8:03 PM
This report is such an excellent breakdown of why polls were so off in 2025 bc modeling decisions undersampling youth voters & Dems, in spite of current data, engineered bad polls overestimating Republicans
And pollsters are continuing the same mistake now in 2026
And pollsters are continuing the same mistake now in 2026