Lucy D’Agostino McGowan
LightNews LightNews
banner
lucystats.bsky.social
Lucy D’Agostino McGowan
@lucystats.bsky.social
5.4K followers 160 following 240 posts
Biostatistician • Associate Prof @ Wake Forest University • former postdoc @ Hopkins Biostat • PhD @ Vandy Biostat • 🎙 Casual Inference • lucymcgowan.com
Posts Replies Media Videos
lucystats.bsky.social
Saw Jen Tullock’s amazing show and was so enamored throughout that I snapped not a single photo, but got to chat after for a bit with the brilliant @hamburgerwaiter.bsky.social …we did discuss elevator pitches 👀😅
November 16, 2025 at 10:37 PM
lucystats.bsky.social
So proud of @ashleymullan.bsky.social -- featured on the cover of @significancemag.bsky.social! Check out her beautiful interrupted time series plot 😍
October 21, 2025 at 4:13 PM
lucystats.bsky.social
I am SO enjoying celebrating International Day of Women in Statistics & Data Science by attending #IDWSDS2025

My first talk of the day:
The Art of Data Refinement: Severance Analyses
🎞️ Slides: lucymcgowan.github.io/mdr-talk/

(All made in Quarto ♥️)

Here is a sneak peak:
October 14, 2025 at 5:10 PM
lucystats.bsky.social
Some corrected headlines 🙏
MMR and V separately, as is common in this age group, is still recommended and available MMR and V separately, as is common in this age group, is still recommended and available MMR and V separately, as is common in this age group, is still recommended and available
September 19, 2025 at 9:58 PM
lucystats.bsky.social
My extremely talented baby sister was featured in this @redbull.com documentary that was released today!

watch the whole thing here 👇
www.soundsofredbull.com/scoring-chal...

listen to more of Teesa’s music:
open.spotify.com/artist/2CttZ...
www.teesadagostino.com
August 23, 2025 at 1:24 AM
lucystats.bsky.social
Whelp 😂
Trying to run tiny gpt-oss on Mac and failing
August 12, 2025 at 11:24 PM
lucystats.bsky.social
Excited for @davisvaughan.bsky.social & my talk tomorrow in which we will demo a shiny app we ✨ vibe coded ✨ up today, if you're at #user2025 stop by Gross at 10:30a tomorrow.

Sneak peak 👇
August 10, 2025 at 1:53 AM
lucystats.bsky.social
The #JSM2025 roasts getting a reprise at @hadley.nz’s #user2025 keynote 😂
roast your colleagues Jeff Leek’s idea of a wild night is running a bootstrapped simulation — twice. Git doesn’t give Jenny merge conflicts; it opens an issue with a minimal reprex and apologizes. Mine’s rubrics are more detailed than most grant proposals — by the time you finish reading them, the semester’s over. Roger believes in reproducibility — which is great, because his jokes in lecture reproducibly bomb every time
August 8, 2025 at 9:24 PM
lucystats.bsky.social
Huge congratulations to two of my favorite colleagues @sarahlotspeich.bsky.social & Staci Hepler on their SUCH well deserved CWS awards at #JSM2025 👏 👏
August 6, 2025 at 4:47 PM
lucystats.bsky.social
Excuse the terrible annotation from my phone (😂) but I think I would have expected the categorical to look more like this:
Bad depiction of categorical age having constant effects within year
July 8, 2025 at 10:05 PM
lucystats.bsky.social
And here is what happens with mean imputation, the slope is the same:
Plot that shows points with imputed values and the slopes from regression with out without observed values
June 26, 2025 at 1:25 AM
lucystats.bsky.social
Maybe a visual will help, here is what is happening with complete case. Even though X is missing NOT at random, we still recover the correct E[Y|X] when just using the observed Xs because the missingness is not related to Y.
Plot showing points missing (and not) and the same slope in both even though missing is not at random (with respect to X)
June 26, 2025 at 1:22 AM
lucystats.bsky.social
Maybe a simple simulation would help? Here I have Y which has a linear relationship with X such that the true effect is 2.

X is missing not at random such that when X > 0 it is missing.

I get 2 when I fit the model with just the observed Xs after listwise deletion and after mean imputation
``` r n <- 10000 effect <- 2 x <- rnorm(n) y <- effect * x + rnorm(n) x_obs <- ifelse(x > 0, NA, x) ## truth (fully observed) lm(y ~ x) #> #> Call: #> lm(formula = y ~ x) #> #> Coefficients: #> (Intercept) x #> 0.002869 2.012952 ## listwise deletion lm(y ~ x_obs) #> #> Call: #> lm(formula = y ~ x_obs) #> #> Coefficients: #> (Intercept) x_obs #> -0.003714 2.026903 ## mean imputation x_imp <- ifelse(is.na(x_obs), mean(x_obs, na.rm = TRUE), x_obs) lm(y ~ x_imp) #> #> Call: #> lm(formula = y ~ x_imp) #> #> Coefficients: #> (Intercept) x_imp #> 1.639 2.027 ``` <sup>Created on 2025-06-25 with [reprex v2.1.1](https://reprex.tidyverse.org)</sup>
June 26, 2025 at 1:15 AM
lucystats.bsky.social
I’m not even sure about the power! I’ve had trouble coming up with a good sim where it noticeably helps, here’s a quick one I just did (on my phone so there are probably bugs 😂) but complete case also has smaller standard errors!
sim <- function() { n <- 100 x1 <- rnorm(n) x2 <- rnorm(n) y <- x1 + x2 + rnorm(n) x1_obs <- ifelse(x1 > 0, NA, x1) x2_obs <- ifelse(x2 > 0, NA, x2) d <- data.frame(y, x1_obs, x2_obs) cc <- broom::tidy(lm(y ~ x1_obs + x2_obs, data = d)) m <- mice::mice(d, print = F, method = "norm") imp <- broom::tidy(mice::pool(with(m, lm(y ~ x1_obs + x2_obs))), conf.int = T) data.frame( cc = cc[2, "std.error"], imp = imp[2, "std.error"] ) } o <- purrr::map_df(1:100, ~sim()) set.seed(1) mean(o$std.err) #> [1] 0.3642946 mean(o$imp) #> [1] 0.3741719
June 25, 2025 at 3:19 AM
lucystats.bsky.social
I am so excited about the slate of amazing tutorials we will have at useR this year! It’s included in your registration, check them out!
| Workshop | Instructors | |---------|------------| | **Causal Machine Learning in R** | Malcolm Barrett | | **Debugging Tools for Functions in R** | E. David Aja & Shannon Pileggi | | **From Model to Meaning: Using the marginaleffects package** | Vincent Arel-Bundock | | **Tidy manipulation of genomic data** | Justin Landis & Michael Love | | **Complex Survey Data Analysis with srvyr and survey** | Rebecca Powell, Isabella Velásquez & Stephanie Zimmer | | **Getting Started with Positron: A Next-Generation IDE** | Jennifer Bryan & Julia Silge | | **R You Out of Memory Again? Level Up with Arrow and DuckDB** | Elyse Armstrong, Jeanne McClure & Sheila Saia | | **Teaching statistics and data science with R and GitHub** | Elijah Meyer & Maria Tackett |
May 20, 2025 at 11:11 PM
lucystats.bsky.social
🎓 🤰 2024 graduation ➡️
👶 2025 graduation
Pregnant graduation photo Baby at graduation photo
May 19, 2025 at 6:10 PM
lucystats.bsky.social
Happy Graduation Weekend from the @wakeforeststats.bsky.social faculty!!
May 18, 2025 at 9:39 PM
lucystats.bsky.social
Some of the causal crew (+ @sarahlotspeich.bsky.social & Staci!) congratulations! 🍾
May 17, 2025 at 10:20 PM
lucystats.bsky.social
😏 @libbyheeren.bsky.social reminding me of the time grad student Lucy snuck a stats joke into Ron Gilbert’s Thimbleweed Park
Ron Gilbert’s Thimbleweed Park (Library, 3rd floor, looking at the book I put in) Book titled Basu’s Theorem by Lucy D’Agostino On a bright summer's day, a young lad, Exbar, decided to embark on an adventure. "I'm feeling complete and sufficient," thought Exbar, "I'm goin' fishing!" he exclaimed. So off he went to collect his fishing gear and begin his journey. Exbar was wandering along a path to the river, when he encountered a snake. This snake looked quite peculiar, almost like an s squared. "Hi. I'm Ann. Ann Ssesssillary," the oddly shaped snake hissed. "I can help you fishhhhhhh," she exclaimed. Feeling complete and sufficient, Exbar confidently replied, "No thanks, Ann. 1 don't need YOU! In fact, I'm independent!" Exbar continued on to the river, where he promptly caught the biggest fish you've ever seen, "Ah! A BASS!" he cried. "Oooo!" chimed in the onlookers. And thus, Basu's theorem: any complete sufficient statistic is independent of any ancillary statistic, such as the mean and sample variance of a normal distribution.
April 30, 2025 at 7:19 PM
lucystats.bsky.social
And our little kid sizes Pepper’s ghost for everyone’s personal hologram, courtesy of a plastic dessert cup and overhead transparency (s/o to Joshua Ellingson for the design!)
March 30, 2025 at 12:27 PM
lucystats.bsky.social
Here’s our lenticular coloring page 😅
March 30, 2025 at 12:27 PM
lucystats.bsky.social
I made these posters with figures from Disney patents to explain the science behind how the projection systems work 🤓
Pepper’s ghost poster with patent Projection patent for Disney
March 30, 2025 at 12:27 PM
lucystats.bsky.social
Ok this was too fun, had some ghosts hanging out on our couch, made lenticular pictures, and made kid-sized Pepper’s ghost contraptions 🙈 the kids loved it and now I need to share with all of my nerd friends 🤗
March 30, 2025 at 12:27 PM
lucystats.bsky.social
Throwing a Pepper’s Ghost themed party for K2 tomorrow, not sure who is more excited 🙈😅
March 29, 2025 at 2:23 AM
lucystats.bsky.social
👀
Plot looking at average pitch for sample of Helly and Helena lines — episode 9 seems to match Helena 👀
March 20, 2025 at 2:20 AM