I don't get how they are getting these estimates for AtlasIntel, given that this was the last poll they published:

Under much resistance from Joy, I took the basement. It's a nice chill space 😄

The contrast of me giving an interview and @joywilke.bsky.social giving an interview, despite both of us being under the same roof!

👀👀👀👀

This is an interesting illustration of potential bias due to non-ignorable selection: respondents from a political poll recruited by ads on Instagram reporting that their main source of information about politics is Instagram.

Oh, most likely it's not entirely nonresponse bias, but it's very likely a factor.

Same mechanism: when the news cycle is bad for one of the sides, their voters tend to answer surveys at lower rates. We can observe that pattern in surveys using the voter file, where we have (modeled) party id.

Also because of partisan nonresponse bias

Partisan nonresponse bias in a nutshell

A couple of examples of what could happen to the betting odds on these types of platforms if there were no election polls:

4) Usar a PNAD Continua mais recentes para obter as distribuição populacionais para o controle da amostra

Especificamente para pesquisas eleitorais:
5) Usar algum modelo de likely voter
6) Separar indecisos de Brancos e Nulos na questão de intenção de voto

1) Presencial domicilar > Presencial por ponto de fluxo > Telefônica > Web
2) Cotas + ponderação > Só ponderação > Só cotas
3) Além das variáveis demográficas de praxe, o ideal também seria controlar (por cotas ou ponderação) por voto na eleição anterior e renda domiciliar
+

Às vezes me perguntam quais considero os melhores institutos de pesquisa no Brasil.
Como metodologista e estatístico, prefiro me concentrar nas melhores práticas metodológicas. Então fica aqui uma colinha para o que considero as melhores prática metodológicas em pesquisas de opinião no Brasil: 🧵+

Esse é o modelo de negócio deles: fazem um monte de pesquisas eleitorais para depois pegar aquelas que eles "acertam" mais e usar como propaganda sobre como eles são os mais precisos do mercado. Um modelo de negócio bem babaca, se você me perguntar...

Eu estou, só que eu ando um pouco sumido daqui hahaha

Rough, must have been quite a surprised! 🥵🔥

Me at looking at the logos of Michigan, Oklahoma and Eastern during the gymnastics meet: "Margin Of Error" 😜

Love surveys? Work with the best!

Careers - BlueLabs We’re always looking to add new talent to our growing team of strategists, data scientists, and software engineers who operate at the forefront of data

Love surveys? We're hiring!

We're looking for new members of the polling team at @bluelabs.bsky.social. Two roles are focused on execution of day-to-day polling operations (Polling Manager, Lead Polling Manager) while the Survey Scientist is an R&D role!

Share & apply!
bluelabs.com/careers/

Huge loss to the public opinion field!

One of my many pet peeves is people who denote survey sample sizes by N.
In surveys statistics, because we are dealing with finite population inferences, we reserve the notation N for population size and use n for sample size.

Outdoor photo of the Convergence sculpture at ISR during a snowfall; a white text box reads “INSTITUTE FOR SOCIAL RESEARCH, Snow Flurries on Convergence.”

Interior view of ISR's atrium with hanging plants and glowing pendant lights; white text box reads “INSTITUTE FOR SOCIAL RESEARCH, Atrium Glow.”

Exterior photo of Perry Building with clear blue sky and red brick facade; white text box reads “INSTITUTE FOR SOCIAL RESEARCH, Perry Brick Red.”

Scenic rooftop view from ISR’s 6th floor showing a blue flag with a yellow M flying above Ann Arbor rooftops; white text box reads “INSTITUTE FOR SOCIAL RESEARCH, View From the 6th Floor.”

The Shades of ISR Pantone ™️ palette 💙 💛

That's one Meng's point in his 2018 paper, isn't it?

Oh yeah, for sure, that's quite clear there. My point is that, as a general rule, statistically speaking, it just doesn't make sense to me, despite the empirical evidence. Bias in surveys doesn't typically scale with sample size, which is what the rule suggests.

I always thought this rule that the total error is 2x reported MoE nonsensical, because bias in surveys doesn’t typically scale with sample size.
In fact, we usually see the opposite: smaller, more carefully designed prob samples tend to present smaller total error than large nonprob samples.

We Are SRC YouTube video by Survey Research Center, University of Michigan

The Survey Research Center at the Institute for Social Research of the University of Michigan turns 80 years old this year. Very proud to be part of this history!
youtu.be/giSsP7o4PZg?...

E o mais importante, a descrição metodológica da pesquisa:

Cenários de 2° turno: +

Cenários de 1° turno com Michelle: +