Our causal inference summer school, Utrecht 6 - 10 July (€875 for academics) is open for registration!
Last year we added a new day focused entirely on designing a causal research project from scratch in groups - looking forward to doing this again this year!
utrechtsummerschool.nl/courses/heal...
After 5 years of data collection, our WARN-D machine learning competition to forecast depression onset is now LIVE! We hope many of you will participate—we have incredibly rich data.
If you share a single thing of my lab this year, please make it this competition.
eiko-fried.com/warn-d-machi...
So much stupid stuff happens when your institutions are motivated by the haunting fear that someone, somewhere is getting something they don't deserve.
It has become received wisdom in Brussels and Washington that there is a new “euro-sclerosis”: that the EU economy is lagging the US
This view is wrong
A little primer on the measurement of productivity – and why reports of the economic death of Europe are greatly exaggerated🧵
NEW PAPER
The use of explainable AI in healthcare evaluated using the well known Explain, Predict and Describe taxonomy by Galit Shmueli
link.springer.com/article/10.1...
Hello, Bluesky! We are ODISSEI Social Data Science Team, also known as SoDa. This is our new official Bluesky account and we would love to interact more with you. 👋
We have a mission of 𝗵𝗲𝗹𝗽𝗶𝗻𝗴 𝗿𝗲𝘀𝗲𝗮𝗿𝗰𝗵𝗲𝗿𝘀 𝘄𝗶𝘁𝗵 𝗱𝗮𝘁𝗮 𝗶𝗻𝘁𝗲𝗻𝘀𝗶𝘃𝗲 𝗮𝗻𝗱 𝗰𝗼𝗺𝗽𝘂𝘁𝗮𝘁𝗶𝗼𝗻𝗮𝗹 𝗿𝗲𝘀𝗲𝗮𝗿𝗰𝗵. 🤝
Dito isto, o que si parece é que ningún deles participará do goberno.
Estase dando así moito esta información, pero a realidade é máis complicada. Hai 3 partidos de extrema dereita no parlamento. O máis grande deixouse un tercio dos votos, pero os outros dous subiron moito. De feito ,a extrema dereita ten 1 escano máis agora dos que tiña antes das elecioón.
Onte houbo eleccións en 🇳🇱 e no fragmentado escenario político destaca un partido pola súa estabilidade.
O SGP que representa as facción máis ortodoxas e conservadoras do calvinismo concentradas no cinto bíblico do país obtivo, como é habitual, 3 deputados.
España ya está a la cabeza de Europa en porcentaje de estudiantes en universidades privadas. Solo le superan tres países minúsculos (Chipre, Malta y Montenegro) y dos del Este (Hungría y Polonia).
A nadie parece preocuparle la transformación estructural que esta tendencia supone.
In 1916 the BMJ published an article about the work done by James Shearer, an American physician working in the British Army as a sergeant (because he had no British qualification). He had described a "delineator" which was better than x rays for portraying gunshot wounds. This caused a sensation and a lot of interest — but on investigation the work was found to have been invented. The BMJ published a retraction, but Shearer was tried by court martial and sentenced to death by firing squad.
Next time an institution tells you how seriously it takes research misconduct, ask them if it's *this* seriously. www.bmj.com/content/297/...
Looking forward to the development of fourthly robust, fifthly robust and why not twenty-seventhly robust estimators to keep up with
![Image of R code. To reproduce:
library(ggplot2)
library(dplyr)
library(mice, warn.conflicts = FALSE)
imp <- mice(nhanes, m = 5, maxit = 5, seed = 1,
ignore = rep(c(FALSE, TRUE), c(20, 5)),
print = FALSE)
impdats <- complete(imp, "all")
train <- lapply(impdats, function(dat) subset(dat, !imp$ignore))
test <- lapply(impdats, function(dat) subset(dat, imp$ignore))
fits <- lapply(train, function(dat) lm(age ~ bmi + hyp + chl, data = dat))
preds <- predict_mi(object = fits, newdata = test, pool = TRUE, interval = "prediction")
preds
preds %>%
as.data.frame() %>%
mutate(case = 1:nrow(preds),
y = test[[1]]$age) %>%
ggplot(aes(x = fit, y = case, col = rowSums(is.na(nhanes[imp$ignore,]))>0)) +
geom_point() +
geom_errorbar(aes(xmin = lwr, xmax = upr)) +
theme_minimal() +
scale_color_manual(values = mice::mdc(1:2), labels = c("observed", "missing")) +
theme(legend.title = element_blank(),
legend.position = "bottom") +
labs(x = "prediction",
title = "Pooled prediction intervals")](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:omgr6y6c62b4qar7fganjtdy/bafkreiesbac4qo25mh5rkegnw4ntcvbwedvpzqms4zn7xiothhzlgchxcm)
Image of R code. To reproduce: library(ggplot2) library(dplyr) library(mice, warn.conflicts = FALSE) imp <- mice(nhanes, m = 5, maxit = 5, seed = 1, ignore = rep(c(FALSE, TRUE), c(20, 5)), print = FALSE) impdats <- complete(imp, "all") train <- lapply(impdats, function(dat) subset(dat, !imp$ignore)) test <- lapply(impdats, function(dat) subset(dat, imp$ignore)) fits <- lapply(train, function(dat) lm(age ~ bmi + hyp + chl, data = dat)) preds <- predict_mi(object = fits, newdata = test, pool = TRUE, interval = "prediction") preds preds %>% as.data.frame() %>% mutate(case = 1:nrow(preds), y = test[[1]]$age) %>% ggplot(aes(x = fit, y = case, col = rowSums(is.na(nhanes[imp$ignore,]))>0)) + geom_point() + geom_errorbar(aes(xmin = lwr, xmax = upr)) + theme_minimal() + scale_color_manual(values = mice::mdc(1:2), labels = c("observed", "missing")) + theme(legend.title = element_blank(), legend.position = "bottom") + labs(x = "prediction", title = "Pooled prediction intervals")
Cool stuff!
Florian van Leeuwen and I implemented a prediction function in the #mice package that allows the incorporation of missing data uncertainty in a prediction interval.
The `predict_mi()` function is available in the current development version: github.com/amices/mice
#Rstats #statsky
This only happens to you once
Haha. Glad we are on the same wavelength.
Hope it’s not too late to recommend Infinite Powers, by @stevenstrogatz.com . It’s fun and very well written!
A lot of the Target Trial Emulation framework is 'just' translating ideas from causal inference with observational data into ideas that are easier to digest and think through, even if you already know the hard stuff.
Now it's time to do some studying and apply these ideas to my own research!
2. Communication is key in science. Just Robins 'g-formula' would conceptually cover a lot of what we did. But I doubt anybody who is presented with just the g-formula would be able to think clearly about their causal question in terms of treatment strategies, postrandomization events, and so forth.
Some thoughts after the course:
1. It is very important to think about treatment strategies and how they develop over time. Not commonly done, but often the inference of most interest in trials is about time-varying treatment strategies, not interventions at one time-point.
Last week I was lucky enough to participate in the Target Trial Emulation course at @causalab.bsky.social. It was a really nice experience! The instructors were excellent, as were the teaching materials, and they covered a great deal of content. I would really recommend it.
Looks good. Thanks!
Can I ask, what would a good 'second' book be in your opinion?
"Have you thought about applying the @opensafely model, for data privacy and efficiency, to non-health data?"
YES WE HAVE
Behold the new... OpenSAFELY-Schools!!
schools.opensafely.org
Are you interested in improving the #interpretability, #robustness and #safety of current AI systems with #causality and #RL?
Apply to our PhD position in Amsterdam 🚲🌷🇳🇱
Deadline: June 15
Nice Thom!
Very happy to announce that the R-package `densityratio` is on CRAN! It implements non-parametric distribution comparison through density ratio estimation, which is useful for sample selection bias adjustment, two-sample testing and more!
See thomvolker.github.io/densityratio for vignettes and info!
Still some spots available in our summer school on all things causal inference, 7-11 July in Utrecht! Discounts for those working in universities and non-profits, and affordable accommodation offered by @utrechtuniversity.bsky.social summer school!
HIRING!
We offer 3 fully funded PhD positions to work on research methodology with experts in a great medical research environment.
More info: www.linkedin.com/feed/update/...
