INSERM Researcher in biostatistics & #rstats enthusiast - views are my own
Joined June 2012
- Tweets2.7K
- Following562
- Followers288
- Likes4.8K
With K Ba, @rodthiebaut & @denis_agniel we noticed @jsb_ucla et al applied Wilcoxon on raw counts while all other methods used normalized counts accounting for library size. When applying all methods on the same (normalized) data, Wilcoxon also gives exaggerated false positives.
Boris Hejblum (@borishej.bsky.social) retweeted
🌎 Excited to be heading to beautiful Portugal for my first International Symposium on Nonparametric Statistics! I'll be talking alongside @borishej, Yanyuan Ma, and @mslaylap about nonparametric methods to take advantage of auxiliary data in health settings! 🏥
Boris Hejblum (@borishej.bsky.social) retweeted
RNAseq counting tools are not perfect. I simulated 240 GTEx samples to test multiple tools. Below I show the difference between actual and estimated counts for each simulated sample.
But, what is causing this? And will it affect differential expression? (1/7) #Bioinformatics
Boris Hejblum (@borishej.bsky.social) retweeted
What is happening in single-cell genomics?! A new paper from the Karasuyama lab at @TMDUniversity (nature.com/articles/s41467-0…) shows three different RNA velocity / pseudotime plots, no pair of which agree, and then merrily goes on to tell some story. 👀
Boris Hejblum (@borishej.bsky.social) retweeted
Find all the fun things here!
r-causal.github.io/r-causal-…
Boris Hejblum (@borishej.bsky.social) retweeted
In a recent preprint with @GorinGennady (biorxiv.org/content/10.1101/…) we provide a quantitative answer to to this question, namely what information about variance (among cells in a cell type, or more generally many cell types) does a UMAP provide? A short🧵1/
Boris Hejblum (@borishej.bsky.social) retweeted
The number of single-cell RNA-seq studies with analyses based on the closed-source CellRanger (>=3) counting/quantification tool is truly astounding and quite depressing given the number (and quality) of the available open-source and transparent alternatives 😪...
Boris Hejblum (@borishej.bsky.social) retweeted
🎙️ On this weeks episode we talk about a “Causal Quartet” a set of four datasets generated under different mechanisms, all with the same statistical summaries (including visualizations!) but different true causal effects
(Plus a chat about M-bias!)
ALT 4 DAGS displaying a collider, confounder, mediator, and m-bias
ALT Four scatter plots for each of the four datasets plotting the exposure vs the outcome — the four look indistinguishable all with a slope of 1
M-Bias: Much Ado About Nothing? | Season 4 Episode 10 casualinfer.libsyn.com/m-bia…
Boris Hejblum (@borishej.bsky.social) retweeted
Follow-up thread on potential approaches by our group and many others towards addressing the
REPLICATION CRISIS IN METHODOLIGAL/COMPUTATIONAL SCIENCE discussed in
nitter.gq/BoulesteixLaure/status…
1/n
Boris Hejblum (@borishej.bsky.social) retweeted
Boris Hejblum (@borishej.bsky.social) retweeted
Oh hi, I wrote a thing for @InstMathStat
Full text here: imstat.org/2023/03/31/writte…
This Post is from an account that no longer exists. Learn more
Boris Hejblum (@borishej.bsky.social) retweeted
For a non-technical overview of this non-replicability written in magazine style with @sabineIk11 @HeidiBaya and A. Charlton, see our article in @Significance : rss.onlinelibrary.wiley.com/…
4/n
🐌 Frustrated with the speed of your #rstats code ? 🧑💻
🚀 "HeavyR: Fast & efficient R code made easy", our course with @_rgenuer_ is next June (2.5 days) 😎
🧑🎓 Register here: bit.ly/3QzjgmV 🥇
ℹ️ more info: [email protected]📩
#datascience #statistics
Boris Hejblum (@borishej.bsky.social) retweeted
The Rdatasets website now hosts over 2000 datasets in CSV format. All data are free and documented. Perfect for teaching or if you're just looking for some data to play with. vincentarelbundock.github.io… #RStats #StatsTwitter #EpiTwitter #EconTwitter #PolisciTwitter
Boris Hejblum (@borishej.bsky.social) retweeted
It is silly to try to separate data analytical methods into machine learning, AI and statistics. They cannot be separated from a historical point of view, from how they are used, for what they are used and by whom they are used
Boris Hejblum (@borishej.bsky.social) retweeted
Publication ce soir dans @NEJM de l'essai PREVAC, large essai clinique randomisé en Afrique de l’ouest évaluant la sécurité et l’immunogenicité de 3 schémas vaccinaux contre Ebola.
Un succès grâce à un gros travail d’équipe et une large collaboration internationale ! 💪 – at Aubervilliers, France
In two randomized trials of Ebola vaccines, safety and immune responses in adults and children were assessed for three regimens with two different vaccines. All three induced immune responses lasting at least 1 year. nej.md/3FMnzYk
Boris Hejblum (@borishej.bsky.social) retweeted
Doing #SingleCell #RNAseq? Ever wonder if all those clusters are real? Turns out most feature selection & clustering pipelines can't tell when there's only 1 cluster! But I found a solution! 🧵👇
