A new paper I worked on is out in Justice Quarterly! I won't speak on the substantive nature of the paper as I worked solely as the methodologist, but I developed a new matching method not otherwise described in the literature, and I want to tell you about it! #statssky #casualsky

opossum my possum

I recently updated my #Rstats package {optweight} for the first time in 6 years(!), and I want to give it the attention it deserves with an announcement and mini thread. In short, {optweight} uses optimization to estimate balancing weights in observational studies. #episky #statssky #causalsky 1/

I could have sworn I created this before on our Previous Parish, but couldn't find it so made it fresh. I present: OUR BLESSED mixed models // THEIR BARBAROUS fixed effects

cowboy depop

I'm teaching this workshop in two weeks! It's going to be great!

These course notes on nonparametric regression (including kernel density estimation) by Eduardo García Portugués are *fantastic*. So clear, with great visuals and clear code.

Looking to strengthen your #causalinference skills? Join @noahgreifer.bsky.social on Aug. 12-15 for "Causal Inference in R Using MatchIt and WeightIt" to gain the skills to apply these #Rstats packages to estimate and interpret treatment effects.

A nice paper by Mark Ratkovic on double machine learning (DML) oriented toward political scientists, with some advancements to accommodate interference and clustering. Definitely a worthwhile read, especially if the foundational papers on this topic intimidate you (as they do me).

A nice recent article on why you should abandon hazard ratios. #statssky #episky

Version 0.5.0 of #Rstats {fwb} is out! {fwb} implements the fractional weighted bootstrap (aka Bayesian bootstrap), which involves repeatedly sampling sets of weights and computing the quantity of interest in each weighted sample. It contains drop-ins for boot::boot() and sandwich::vcovBS().

Randomly obsessed with simultaneous (uniform) inference. Feel free to ask about it. A mini thread on this follows. 🧵 Recommended reading: #statssky #econsky #episky

An idea for expedited reviews: Show me your abstract and I'll send you the @dingdingpeng.the100.ci paper you should've read before you started your project.

Begging you to rank Zohran Mamdani for NYC mayor, and please DO NOT rank Cuomo.

You know I had something to say.

Bayesian Additive Regression Trees are fucking cool.

about that: www.sciencedirect.com/science/arti...

Begging you to rank Zohran Mamdani for NYC mayor, and please DO NOT rank Cuomo.

I'm trying to find someone to teach a short course helping SAS users (specifically) transition to #Rstats. If you are such a person or are willing to recommend someone, my DMs are open!

Good news, Harvard renewed my contract (for 6 months) so I remain employed and will continue to do my best to write great R packages and support applied research for Harvard and MIT affiliates :) Thank you so much for the support in a vulnerable moment!

github.com/kylebutts/vs... If you are using vscode and stata, you should try out my extension. It uses interactive window which let's you write in a `.do` file but get a notebook type experience.

Starting to look like I might not be able to work at Harvard anymore due to recent funding cuts. If you know of any open statistical consulting positions that support remote work or are NYC-based, please reach out! 😅

In October I'll be giving a 2-day workshop at @stathorizons.bsky.social about how to make beautiful websites with #QuartoPub! Learn how to make and deploy personal websites, research websites, an course websites all with minimal HTML/CSS

📣 July 22-25 I’ll be teaching an online short course on how to read and understand statistics commonly found in medical literature — tell your clinician friends! @stathorizons.bsky.social statisticalhorizons.com/seminars/und...

Working on an #rstats package for continuous treatments… any clever name ideas? (Preferably not ending in “It” 😑)

Starting to feel like "don't look at the coefficients, just calculate whatever metric is relevant to your research question" is a highly underappreciated stats hack and also I may have to get myself a marginaleffects T-shirt.

I love everything @vincentab.bsky.social does, but I truly believe in {marginaleffects} and I'm so glad he got the opportunity to tell more people about it on this episode of @casualinfer.bsky.social. I'm honored to have been a part of the package and get a nice shoutout :)

Thinking about odds ratios... An odds is a ratio of events to non-events. For example, if the event is survival, the odds of survival is the number of survivors per death. If the event is getting a disease, the odds is the number of diseased individuals per healthy individual.

Resisting the urge to restart the negative binomial vs. Poisson wars...

Really nice new paper by Jingyu Zhang, Oliver Lüdtke, and Alexander Robitzsch on the performance of doubly robust estimators of the ATE. A great example of clear writing and reporting, useful visualization through tables, and a review of modern literature. osf.io/5uj2f_v2

I loved being a guest on the Casual Inference Podcast with @lucystats.bsky.social and @epiellie.bsky.social, two of my science heroes! Listen to hear us talk about overlap weighting, balance, CBPS, M-estimation, managing trade-offs, and science communication! #episky #statssky #causalsky

A reminder not to include your ID variable when doing multiple imputation :) Seen this too many times

A reminder that if you are at Harvard or MIT (in any department, including affiliated hospitals and centers), you can get free statistical consulting services from me and my teammates through IQSS's Data Science Services. Details here:

The Essential Role of the Phrase "The Essential Role" in Giving a Paper The Role of "Essential"

In a few weeks I’ll be giving a 2-day workshop on the interpretation of statistical models in #Rstats. We’ll be talking a lot about the {marginaleffects} 📦. I ❤️ teaching that class! The participants are super fun and our conversations are always interesting. Let me know if you have any questions!

When I tell people I'm obsessed with sudoku, I always have to say "not the NYT kind; artisanal sudoku!" Today I solved a really beautiful and elegant one that I think isn't too hard and is a good example of how artisanal sudoku differs greatly from the standard kind. sudokupad.app/gxcfo7dg0y

I'm so excited to announce that I'll be teaching "Causal Inference in R Using MatchIt and WeightIt" with @stathorizons.bsky.social! In this 4-day workshop (Aug 12 - Aug 15), I'll explain the basics of causal inference, matching, and weighting, using my R packages {MatchIt} and {WeightIt}.

thomas "the tank" pynchon

Of course I decided to learn surgery by reading some blog posts, and the documentation for the {suRgery} package was good enough to get the job done (though I typically just use the defaults and hope for the best). So writing the primer was the least I could do for the medical community.

Trying various combinations of eval(), quote(), alist(), and substitute() until whatever blasphemy I’m trying to accomplish works #rstats

Grant me the strength to understand Riesz representers

The covariate balancing propensity score (CBPS) is a method of estimating PS weights that satisfy the balance constraint A*w*X - (1 - A)*w*X = 0 where A is 0/1 tx, w are the weights (a function of the PS depending on the estimand), and X is a matrix of covariates and an intercept. 🧵 #statsky

Made my own Connections puzzle for those as into causal inference and statistics as I am :) Give it a try and let me know what you think! (Slightly NSFW clues) #statsky #causalsky #episky connectionsplus.io/game/x8nqcT

Made my own Connections puzzle for those as into causal inference and statistics as I am :) Give it a try and let me know what you think! (Slightly NSFW clues) #statsky #causalsky #episky connectionsplus.io/game/x8nqcT

If you have a sample of size n with mean M and you want to compute the new mean M* after adding an additional observation x, a numerically stable formula is M* = M + (x - M)/(n + 1) This minimizes inaccuracies due to dividing large numbers because (x - M) tends to be small.

New paper for holiday reading! @isager.bsky.social and I provide an introduction to three-sided testing, a framework for testing estimates' practical significance. We offer a tutorial, Shiny app, + commands/code in #Rstats, #Jamovi, + #Stata. 1/9 osf.io/preprints/psyarxiv/8y925 #EconSky #PsychSky

If you have an issue where Dropbox is constantly syncing while @posit.co #RStudio is open, one solution is to run the following code in Terminal: find "Dropbox" -type d -name ".Rproj.user" -print0 | while IFS= read -r -d $'\0' folder; do xattr -w com.dropbox.ignored 1 "$folder"; done #rstats

I’m done posting controversial stuff. Helpful posts from now on.

A point estimate? You mean a 0% confidence interval?