blog post: toy example for energy balancing weights How do energy balancing weights (used now by the NYT/Siena Poll) handle unsampled population groups ? Let's work thru a toy example and compare to Poststratification, Raking, and MRP.

blog post: Big Changes in the Times/Siena Poll 2 changes to their survey weights: 1. new weighting variable: support score 2. new weighting method: energy balancing

blog post: perfect collinearity in the sample but not in the population Two variables are perfectly collinear in your sample, so you drop one. You use your model to predict in the population. What can go wrong ? Let's talk thru a Census Bureau toy example from BDA2.

blog post: using MRP in later analyses (pride edition) happy pride ! 🌈 @jeffreylax.bsky.social & Phillips 2009 used MRP to estimate state-level public opinion about policies affecting gays and lesbians. They then use this as a predictor of whether the state adopts the policies.

blog post: should MRP workflow include LOCO-CV ? Individual-level loss orders models differently than the population-level loss we want judging MRP. To get population-level loss, use out-of-sample classical poststratification to compare with MRP. How to split data ? LOCO = leave one cell out.

blog post: it is (still) the people Survey Statistics blog series' 1st birthday 🥳 Andrew Gelman's 60-ish Birthday 🥳 and NYT weights with synthetic past vote 🗳️

blog post: double-plus robustness Meng (2022): GREG is not only “double robust” (consistent if either the outcome model or response model are correct), but “double-plus robust” (consistent if what is left of the outcome model and response model are uncorrelated).

blog post: GREG GREG is Generalized REGression estimator. We can think of it either as: 1. Adjusting an estimate based on the model with a Horvitz-Thompson estimate of the error, or 2. On the flip side, adjusting the Horvitz-Thompson estimate with the model.

blog post: relevant alternatives ? We saw that the multinomial logit model implies independence from irrelevant alternatives (IIA). Let’s expand the model to include choice set C within the logits f(X_ic,C), allowing for non-IIA.

Survey Statistics: Blue Rose Research is (still) hiring ! statmodeling.stat.columbia.edu/2026/05/05/s...

blog post: work with us at Blue Rose ! use cutting edge statistics, machine learning, and engineering to study public opinion, forecast elections, and advise Democrats.

blog post: exploded logit ! a common choice model is multinomial logit. this model implies that rankings follow an exploded logit !

blog post: irrelevant alternatives ? a common choice model is multinomial logit. this model implies Independence of Irrelevant Alternatives (IIA), e.g. the ratio of Left-vs-Right preference is the same in round 1 as in the runoff.

blog post: improving with structure We’ve met Mr. P (Multilevel Regression and Poststratification). We’ve met Mrs. P (Multilevel Regression with Synthetic Poststratification). Now let’s meet Ms. P (Multilevel Structured regression with Poststratification).

blog post: design-based cross validation how to split train and test sets to respect survey design ? what lessons carry over to nonprobability samples ?

blog post: Individualism and the CV Noise Problem Politically meaningful differences among models can be swamped by cross-validation noise.

blog post: individualism doesn't work (even when weighted) individual-level loss (even weighted to the population) orders models differently than the population-level loss of interest to folks using MRP

blog post: work with us at Blue Rose ! use cutting edge statistics, machine learning, and engineering to study public opinion, forecast elections, and advise Democrats.

blog post: sampling-weighted loss we use sampling weights to estimate a population mean E(Y). what about to estimate a conditional mean E(Y|X) ? the best-fit model in the sample may not be the best-fit model in the population.

blog post: sampling to assess data quality @bhedtgauthier.bsky.social et al. (2012) used sampling to assess and improve data quality in Malawi

blog post: Gallup's Presidential Approval Ratings Gallup will no longer track presidential approval after 88 years Let's look at their sampling, mode, and weighting (still used for other survey questions)

blog post: more on recalled vote we've talked about measurement error in recalled vote in the US. how does this change in multiparty states ?

We recently tested ~ a dozen public statements from a diverse set of Democratic elected officials on the murder of Renee Good and this was the top testing one

blog post: 5 flavors of calibration 2 from survey statistics 1 from machine learning 2 from Gelman et al.'s workflow article

blog post: Total Margin of Error (Part II) For election polls from 1998 to 2014 Shirani-Mehr et al. found: margin of error = 2 x (reported margin of error) Let's revisit Meng's “Statistical Paradises and Paradoxes” to understand this more generally.

blog post: Total Margin of Error margin of error = 2 x (reported margin of error) and how much of this error is "bias" vs "variance" ?

blog post: Margin of Error how can we get a poll's margin of error ? let's start with MRP and some simplifying assumptions.

This post has some more discussion of other methods for incorporating known ground-truth margins in an MRP framework, based on some validation exercises in osf.io/preprints/so...

blog post: 4th helpings of the logit shift y_1 = governor vote choice y_2 = abortion proposition vote choice x = demographics You want E(y_2 | county). You have y_1, y_2, x in a survey, x in the population, and E(y_1 | county). @wpmarble.bsky.social and Josh Clinton have ideas !

blog posts: should we adjust for a mismeasured X ? You know the population distribution for X (e.g. vote choice in 2024). But you only have a reported X* in your survey. Should you adjust for it ? Later today: exploring toy examples to see.

blog post: 3rd helpings of the logit shift You have multiple outcomes, but only some have aggregate truth to shift to. How can we calibrate our estimates of p(y_1, y_2 | X) to aggregate data about E[y_1] ? @wpmarble.bsky.social and Josh Clinton have ideas !

blog post: 3 probabilities in Meng 2022 1. (human) design probabilities, e.g. P[R = 1 | stratum] in stratified sampling 2. divine probabilities, e.g. P[R = 1 | anything about a person] where responders follow laws of nature 3. device probabilities, e.g. P[R = 1 | X] modeled

blog post: probability sample = known nonzero probability epsem = equal individual probabilities SRS = equal entire-sample probabilities

Coming soon: our introduction to the politics of human rights 🥳📚 Preorder available here: www.cambridge.org/highereducat... @sabinecarey.bsky.social

blog post: quantity vs quality compare 2 surveys: 1. 100% coverage, but response probability P[R = 1 | Y] differs a lot by Y 2. Only 5% coverage, but P[R = 1 | Y] is roughly constant across Y which would you use ? both ?

new blog post: sampling the sample we’ve focused on estimating means E[Y]. but say Y are openends ("describe how you feel about the candidate") and you want to read thru a few draws from the population, not only survey responders. what should you do ?

blog post: weights and MRP for voters so far we've talked about weights and MRP for E[Y], vote choice in the population overall. but what if you want E[Y | V = 1], vote choice in the population of voters. what are the weights and how do you modify MRP ?

Survey Statistics: continued struggles with equivalent weights statmodeling.stat.columbia.edu/2025/11/04/s...

blog post: Blue Rose Research is hiring ! We are looking for a teammate with expertise in both LLM tools and statistical modeling. Someone who clearly communicates assumptions, results, and uncertainty. With care and kindness.

blog post: individualism doesn't work typical machine learning loss looks at one individual at a time but for MRP, we care about aggregates

blog post: MRPW you've got a survey collected by someone else, and they gave you weights. how can you use those weights in the MRP (Multilevel Regression and Poststratification) ?

blog post: struggles with equivalent weights you've done MRP. someone asks you for survey weights. how to get them ?

blog post: beyond balancing in midterms, voters tend to support the out party for balance do polls still help predict midterms ? yes

The ultimate New York City lover’s treasure hunt is back Friday, Oct. 17 through Sunday, Oct. 19: bit.ly/3IHnHwn

blog post: Fat Bear Week Basu's Bears is a lesson in: 1) using auxiliary information (pre-salmon-feasting weights) 2) how bad an unbiased estimator can be statmodeling.stat.columbia.edu/2025/09/23/s...

blog post: random sampling is not leaving we turned to response instrument Z because random sampling is "dead" but does this method still rely on starting with random sampling ?

blog post: random sampling is not leaving we turned to response instrument Z because random sampling is "dead" but does this method still rely on starting with random sampling ?

blog post on imputation (again): we want E[Y|X] but X can be missing @lucystats.bsky.social @sarahlotspeich.bsky.social @glenmartin.bsky.social @maartenvsmeden.bsky.social et al. say: random imputation should use Y deterministic imputation shouldn't statmodeling.stat.columbia.edu/2025/09/09/s...

I learn from every interaction with Maria. I attended a talk of hers yesterday. I highly recommend reading her work.

blog post: connections between survey statistics and experimental design. split-plot designs are analogous to cluster sampling. blocking is analogous to stratification. featuring an experiment by Arjun Potter and colleagues at NM-AIST !