Christian Testa

I've been reading about missForest today

MissForest—non-parametric missing value imputation for mixed-type data

https://academic.oup.com/bioinformatics/article/28/1/112/219101

https://github.com/stekhoven/missForest

Runs much faster than `{mice}` in my experience, and I like the fewer parametric assumptions.

The above article on missForest is David Stekhoven and Peter Bühlmann's most cited article.

#DataScience #statistics #academia #econometrics #Epidemiology

Is anyone in the US CDC / NIH even working on the Ebola Bundibugyo strain now?

Over a year ago labs these were shuttered

https://www.wired.com/story/hhs-niaid-irf-ebola-disease-research-stop/

We are losing so many years of critical research on infectious diseases that could save countless lives

#InfectiousDisease #Epidemiology #PublicHealth

An enjoyable read I stumbled on

https://www.ecologicalcitizen.net/pdfs/epub-139.pdf

How solarpunk can help us
rewild our lives
Ana Sun and Susan Kaye Quinn

#Solarpunk

Some choice quotes on the topic of instrumental variables:

From Swanson and Hernán 2014

Imbens states we are “limited in the questions we can answer credibly and precisely.” We agree, but there are differences between the questions we can answer and the questions we want answered. Choosing only answerable questions (e.g., identifying the LATE in some settings) can mislead decision-making efforts: our estimates may be misinterpreted as directly relevant to a decision when in fact they are only tangentially related.

From Levis, Kennedy, and Keele 2024

Often in the literature, it is noted that the choice of estimand should be based on a scientific judgement and should reflect the specific research question. However, it is often the case that there is some interplay between the estimand and the plausibility of the identification assumptions. Nowhere is this more true than in an IV analysis.

What do you think? Do you have a favorite instrumental variable?

Tell me it's not rainfall!

https://www.jonathanmellon.com/publication/rain/

I’ve been trying to read more carefully about instrumental variables and make up my mind about when IV arguments are scientifically convincing.

Here's a tension I keep running into:

Should the scientific question alone determine the causal parameter of interest?

Or is it legitimate for the target parameter to reflect an interplay between scientific interest and the identifying assumptions we actually find tenable?

IVs can be difficult to interpret when instruments are weak, who “compliers” are is opaque, exclusion restrictions are debatable, or linear models are used in settings where the true data-generating process may be nonlinear.

On the other hand, when an entire body of (aspirationally causal) literature rests on methods that try to close backdoor paths, IVs offer a genuinely different identification strategy. That seems valuable for evidence triangulation, even if IV analyses have their criticisms.

What do you think? Are you a big IV proponent? Are you an IV critic?

When do you find IV evidence persuasive?

Some literature I've been reading & re-reading:

https://pubmed.ncbi.nlm.nih.gov/16755261/

https://academic.oup.com/ije/article/47/4/1289/3095892

https://pmc.ncbi.nlm.nih.gov/articles/PMC4285626/

https://arxiv.org/abs/2402.09332

https://arxiv.org/abs/2402.05639

#CausalInference #InstrumentalVariables #Econometrics #Statistics #DataScience #HealthPolicy

Very proud to announce the first version of my new R package {nadir} was released today on CRAN!

Check it out --
https://ctesta01.github.io/nadir/

{nadir} is a package for super-learning, which is ensemble weighting candidate machine learners together according to their estimated cross-validated risk on held-out data.

#rstats #software #academia #DataScience #statistics #MachineLearning

https://hsph.harvard.edu/news/trump-administration-freezes-2-2-billion-in-grants-to-harvard/

Today basically every faculty I ran into at the Harvard School of Public Health told me they had their grants cancelled — for work they do towards promoting health and saving lives.

I hope someday soon when you look at academic papers where the usual grant funding disclaimer would have been, that instead you will see “this paper was published in spite of the unlawful termination of grant XXXXXXX.”

I know funds being cut will slow down research and science in the US, but damn if this doesn’t make me want to dig in my heels and work harder for the things I believe in.

My research assistant sure seems fond of generating high quality evidence for treat(ment effect)s 🐾
#dogsofmastodon

I’d like to start a conversation about the #accessibility of math. Someday I aspire to be a professor, and I want to be inclusive to students of a wide range of backgrounds and experiences, but math (esp. advanced math) is full of details that make it very hard for anyone with dyslexia, dysgraphia, challenges with attention span, etc.

I know there’s some things I can do like providing notes / slides in advance so people don’t have to rely on my blackboard handwriting, and providing learning materials in multiple modalities, but what else is needed to make #math a more welcoming subject for students with what I think have traditionally been stigmatized as “learning difficulties”?

The world has spent a long time trying to force everyone to fit their expectations (“sit down, be quiet” ad nauseum) but I dream of a future where the classroom is more genuinely accepting and welcoming to people who experience the world differently than those who fit the mold we’re culturally conditioned to believe is “normal” or neurotypical.

Here’s a specific to get the conversation going: Is web-based math (MathJax / MathML) any better for screen readers than math PDFs? Do any screen readers even work with more advanced math at all??