Raphael Nishimura
@rnishimura.bsky.social
Survey methodologist and statistician at the Survey Research Center, Institute for Social Research, University of Michigan
#survey | #sampling | #statistics | #RStats
#survey | #sampling | #statistics | #RStats
Did you say "representative sample"?
November 7, 2025 at 11:44 PM
Did you say "representative sample"?
A year ago today, AtlasIntel said they would deposit at the Roper Center the raw data from their final polls. It's still not there.
Something tells me they will not be depositing that data file, neither the ones from this week's polls...
Something tells me they will not be depositing that data file, neither the ones from this week's polls...
November 7, 2025 at 8:47 PM
A year ago today, AtlasIntel said they would deposit at the Roper Center the raw data from their final polls. It's still not there.
Something tells me they will not be depositing that data file, neither the ones from this week's polls...
Something tells me they will not be depositing that data file, neither the ones from this week's polls...
Believe or not, the New Jersey poll wasn't the lowest point in US elections for the (former?) "American most accurate pollster", AtlasIntel
Take a look at their final poll in the PNP Primaries in Puerto Rico last year:
Poll: Pierluisi +14
Result: González +9
That's a 23-points difference!
Take a look at their final poll in the PNP Primaries in Puerto Rico last year:
Poll: Pierluisi +14
Result: González +9
That's a 23-points difference!
November 6, 2025 at 3:05 AM
Believe or not, the New Jersey poll wasn't the lowest point in US elections for the (former?) "American most accurate pollster", AtlasIntel
Take a look at their final poll in the PNP Primaries in Puerto Rico last year:
Poll: Pierluisi +14
Result: González +9
That's a 23-points difference!
Take a look at their final poll in the PNP Primaries in Puerto Rico last year:
Poll: Pierluisi +14
Result: González +9
That's a 23-points difference!
l guess NJ is a small demographic subgroup for AtlasIntel standards?
November 5, 2025 at 5:06 AM
l guess NJ is a small demographic subgroup for AtlasIntel standards?
They literally just published a NYC poll two days ago, but sure, why not?
November 3, 2025 at 4:11 AM
They literally just published a NYC poll two days ago, but sure, why not?
LMAO That's not how margins of error work. They reflect random sampling error (variance), not systematic errors, which clearly seems to be the case here, when looking across polls.
Also, saying with a straight face that what's shady is if the crosstab looks okay. C'mon, you are not even trying! 😅
Also, saying with a straight face that what's shady is if the crosstab looks okay. C'mon, you are not even trying! 😅
November 3, 2025 at 1:40 AM
LMAO That's not how margins of error work. They reflect random sampling error (variance), not systematic errors, which clearly seems to be the case here, when looking across polls.
Also, saying with a straight face that what's shady is if the crosstab looks okay. C'mon, you are not even trying! 😅
Also, saying with a straight face that what's shady is if the crosstab looks okay. C'mon, you are not even trying! 😅
By the way, in a week this post will be 1 year old and they haven't posted their data to the Roper Center yet... Something tells me this is not going to happen...
November 1, 2025 at 2:31 AM
By the way, in a week this post will be 1 year old and they haven't posted their data to the Roper Center yet... Something tells me this is not going to happen...
P.S. 2: I wouldn't try to argue with me on this
[meme by my lovely better half, @joywilke.bsky.social 😅]
[meme by my lovely better half, @joywilke.bsky.social 😅]
October 22, 2025 at 2:34 PM
P.S. 2: I wouldn't try to argue with me on this
[meme by my lovely better half, @joywilke.bsky.social 😅]
[meme by my lovely better half, @joywilke.bsky.social 😅]
Oh, I see some new weighting variables here. Cool!
And yes, the methods report is the very first thing I look at a new poll 🤓
And yes, the methods report is the very first thing I look at a new poll 🤓
September 30, 2025 at 3:11 PM
Oh, I see some new weighting variables here. Cool!
And yes, the methods report is the very first thing I look at a new poll 🤓
And yes, the methods report is the very first thing I look at a new poll 🤓
Interesting result on the effect of presenting a "No opinion" option on this Washington Post-Ipsos poll
www.washingtonpost.com/politics/202...
www.washingtonpost.com/politics/202...
September 19, 2025 at 1:46 PM
Interesting result on the effect of presenting a "No opinion" option on this Washington Post-Ipsos poll
www.washingtonpost.com/politics/202...
www.washingtonpost.com/politics/202...
Trump is threatening Brazil with 50% tariffs for no reason whatsoever -- It's Brazil that has a trade deficit with the US...
July 9, 2025 at 11:11 PM
Trump is threatening Brazil with 50% tariffs for no reason whatsoever -- It's Brazil that has a trade deficit with the US...
I wouldn't hold my hopes on that:
"We don't need to set any standards given that anyone who is a member or future member of this organization has already established that they are among the most accurate pollsters in America"
Like, they are not even trying 😅
"We don't need to set any standards given that anyone who is a member or future member of this organization has already established that they are among the most accurate pollsters in America"
Like, they are not even trying 😅
July 8, 2025 at 1:24 PM
I wouldn't hold my hopes on that:
"We don't need to set any standards given that anyone who is a member or future member of this organization has already established that they are among the most accurate pollsters in America"
Like, they are not even trying 😅
"We don't need to set any standards given that anyone who is a member or future member of this organization has already established that they are among the most accurate pollsters in America"
Like, they are not even trying 😅
This has to be a joke, right?
July 8, 2025 at 12:35 PM
This has to be a joke, right?
It's great to call attention to this source of survey error!
Just on a technical note: in survey methods, I think we would call it more generally error of observation or measurement error, under the total survey error framework, per Groves et al (2009):
Just on a technical note: in survey methods, I think we would call it more generally error of observation or measurement error, under the total survey error framework, per Groves et al (2009):
July 6, 2025 at 4:54 PM
It's great to call attention to this source of survey error!
Just on a technical note: in survey methods, I think we would call it more generally error of observation or measurement error, under the total survey error framework, per Groves et al (2009):
Just on a technical note: in survey methods, I think we would call it more generally error of observation or measurement error, under the total survey error framework, per Groves et al (2009):
Since I reposted this yesterday, I feel obliged to also post this important correction to the charts:
June 30, 2025 at 10:02 PM
Since I reposted this yesterday, I feel obliged to also post this important correction to the charts:
Reminder that if these polls are using online opt-in panels, results for young people can be misleading.
www.pewresearch.org/short-reads/...
www.pewresearch.org/short-reads/...
June 13, 2025 at 3:22 PM
Reminder that if these polls are using online opt-in panels, results for young people can be misleading.
www.pewresearch.org/short-reads/...
www.pewresearch.org/short-reads/...
One of my son's toys is the survivorship bias aircraft 😍🤓
June 10, 2025 at 1:06 AM
One of my son's toys is the survivorship bias aircraft 😍🤓
As long you follow Cromwell's rule, Bayes got your back!
June 5, 2025 at 2:28 PM
As long you follow Cromwell's rule, Bayes got your back!
According to the June The Economist/YouGov Poll, 30% of Americans think immigrants shouldn't have the same right to express political views online as U.S. citizens. About half of Republicans/Trump voters shares that view. So much for freedom of speech!
d3nkl3psvxxpe9.cloudfront.net/documents/ec...
d3nkl3psvxxpe9.cloudfront.net/documents/ec...
June 3, 2025 at 5:34 PM
According to the June The Economist/YouGov Poll, 30% of Americans think immigrants shouldn't have the same right to express political views online as U.S. citizens. About half of Republicans/Trump voters shares that view. So much for freedom of speech!
d3nkl3psvxxpe9.cloudfront.net/documents/ec...
d3nkl3psvxxpe9.cloudfront.net/documents/ec...
This is once again to show that betting markets are not oracles or even forecasters per se, they just react to data. Not necessarily a bad thing, that's essentially what Bayesians do, they update priors based on data. But that's all it is.
June 2, 2025 at 1:41 AM
This is once again to show that betting markets are not oracles or even forecasters per se, they just react to data. Not necessarily a bad thing, that's essentially what Bayesians do, they update priors based on data. But that's all it is.
This footnote on the NYT analysis that has been ciculing around tells exactly the pain Connecticut has been causing on those of us using national county-level data 😅😫😭
May 26, 2025 at 1:37 PM
This footnote on the NYT analysis that has been ciculing around tells exactly the pain Connecticut has been causing on those of us using national county-level data 😅😫😭
This is completely infuriating. People work extremely hard and changed their entire lives to move to a whole new country, away from their family and friends, to pursue the highest level of education just to have everything they worked for suddenly taken away from them for no good reason at all...
May 22, 2025 at 7:25 PM
This is completely infuriating. People work extremely hard and changed their entire lives to move to a whole new country, away from their family and friends, to pursue the highest level of education just to have everything they worked for suddenly taken away from them for no good reason at all...
And the author/Silver can't be seriously suggesting that "polls were off" because pollsters dislike Trump!!
Not only that'd be extremely unprofessional, but everyone was extremely concerned about under-estimating Trump once again...If anything, it shows they were letting the data speak by itself.
Not only that'd be extremely unprofessional, but everyone was extremely concerned about under-estimating Trump once again...If anything, it shows they were letting the data speak by itself.
May 19, 2025 at 1:55 PM
And the author/Silver can't be seriously suggesting that "polls were off" because pollsters dislike Trump!!
Not only that'd be extremely unprofessional, but everyone was extremely concerned about under-estimating Trump once again...If anything, it shows they were letting the data speak by itself.
Not only that'd be extremely unprofessional, but everyone was extremely concerned about under-estimating Trump once again...If anything, it shows they were letting the data speak by itself.