Jonathan Bartlett

We are looking forward to hearing @jonathan-bartlett.bsky.social speak on the G-formula for causal inference using synthetic multiple imputation at the July @vicbiostat.bsky.social seminar!

All welcome online Thursday 24th, 4:00pm Aus EST (7:00am UK time).

www.vicbiostat.org.au/event/g-form...

PhD Candidates Causal machine learning – Performance assessment of causal predictive algorithms | LUMC Do you want to work on challenging problems within causal inference and contribute to algorithms that support treatment decisions for individual patients? As PhD candidate causal machine learning at t...

HIRING!

2 PhD openings within the “Safe Causal Inference” consortium with experts from biostatistics, computer science, math, and epidemiology.

You'll develop new methods to evaluate prediction algorithms that take the causal effect of treatments into account.

👉 www.lumc.nl/en/about-lum....

1/ NEW R PACKAGE! For estimating the impact of potential interventions on multiple mediators in countering exposure effects (led by @cttc101.bsky.social)

- Paper👉 tinyurl.com/ye26jsps
- Package👉 tinyurl.com/yuh4kens

Thread shows published examples of how the method can be used! #EpiSky #CausalSky

📣 Calling everyone working in #datascience #biostatistics #clinicaltrials

We’re bringing together experts on target-trial emulation and other frameworks, where we’ll explore the role and potential of observational data for evaluating the effects of interventions

Don’t miss out 🔽
bit.ly/TTE_25

🚨 Next month, we’ll be hosting a one-day event on target-trial emulation and other frameworks, exploring the role and potential of observational data for evaluating the effects of interventions

Open to everyone working in #datascience #biostatistics #clinicaltrials

Get your ticket 🔽
bit.ly/TTE_25

I probably misunderstand, but when you install a package it will install other packages it depends on. And then when you load the package with library() it loads the dependencies likewise.

Prediction under intervention: challenges and trade-offs | LSHTM Causality and prediction are often two separate activities. In particular, prediction can be done in a way that is agnostic to underlying knowledge, mechanism or causal structure. However, it is very

Join us on 10th June (online or in London @lshtm-dash.bsky.social ) to hear from Matthew Sperrin talk about his work on 'Prediction under intervention: challenges and trade-offs'.More details at www.lshtm.ac.uk/newsevents/e...

📆 SAVE THE DATE: 26 June 📆 for our 1-day event on “Target trial emulation and other frameworks: The role and potential of observational data for evaluating effects of interventions”, hosted by the Centre for Data & Statistical Science for Health (DASH) at LSHTM. @lshtm-dash.bsky.social

The estimands framework: a primer on the ICH E9(R1) addendum Estimands can be used in studies of healthcare interventions to clarify the interpretation of treatment effects. The addendum to the ICH E9 harmonised guideline on statistical principles for clinical ...

Indeed. This paper is a good overview of the ICH E9 addendum on estimands on this topic: doi.org/10.1136/bmj-...

Yes it could. These hypothetical estimands do indeed deviate from what I have always interpreted ITT to mean. For me ITT means analyse according to randomised group and look at outcomes irrespective of events such as treatment switch.

The role of post intercurrent event data in the estimation of hypothetical estimands in clinical trials Clinical trial estimands which make use of the so-called hypothetical strategy target the effect of one randomised treatment compared to another in a scenario where the corresponding intercurrent e…

Should data observed after intercurrent events handled by the hypothetical strategy be used in estimation of treatment effects? Rhian Daniel and I investigate... thestatsgeek.com/2025/04/03/t...

'G-formula with multiple imputation for causal inference with incomplete data'. Open access in Statistical Methods in Medical Research. doi.org/10.1177/0962...

Handling multivariable missing data in causal mediation... : Epidemiology miologic studies. However, guidance is lacking on best practice for using multiple imputation when estimating interventional mediation effects, specifically regarding the role of missingness mechanism...

📣 📣NEW PAPER providing guidance on best practice for using multiple imputation when estimating interventional mediation effects, considering missingness mechanism, multiple imputation model specification, & variance estimation
#CausalSky #EpiSky

Read more 👇🏽
journals.lww.com/epidem/abstr...

Powering RCTs for marginal effects with GLMs using prognostic score adjustment In randomized clinical trials (RCTs), the accurate estimation of marginal treatment effects is crucial for determining the efficacy of interventions. Enhancing the statistical power of these analyses ...

New paper! We extend my prior work on prognostic adjustment to work with generalized linear models. This is a nice way to gain power in randomized trials (eg with binary outcomes) by leveraging historical data in a way that does not sacrifice type I error control.

arxiv.org/abs/2503.22284

Sorry. I agree with you! My initial reaction/thinking was that in conditional imputation there are two variables in play, with one only defined in those for whom the first takes a certain value. But as you indicate, you can translate this into a problem with one variable. Thank you!

Probably looking at the example in the vignette will (hopefully!) make it clear.

Not the same I don't think. This is about a situation similar to censoring- you have partial info about the missing values. The smcfcs additions are though for factor variables, where instead of the exact category, you know someone belongs to one among a subset of the categories...

Multiple imputation for coarsened (grouped) factor covariates Missing data are a common problem in statistical analyses. A closely related but slightly different problem is when for an individual in a dataset, although we do not know the exact value of a part…

Imputation of factor variables when you have partial information about some of the missing values. See here for more details thestatsgeek.com/2025/03/27/m...

Demystifying Bayesian meta-analysis for researchers | LSHTM Bayesian models offer a powerful framework for meta-analysis through their flexible and probabilistic treatment of uncertainty.There are several methodological challenges in evidence synthesis,

3rd April, in London and online, come and hear @rlgrant.bsky.social talk about his new book with Gian Luca Di Tanna on Bayesian meta-analysis. Further details at www.lshtm.ac.uk/newsevents/e...

The EuroCIM program is live! Explore the sessions, speakers, and schedule here: www.eurocim.org/program.html. Get ready for an exciting conference! 🎉

I think I sometimes write estimates are biased - they are the realisations of a biased estimator. But I know it's not strictly correct.

It's OK you can name and shame me!

'Missing baseline data when analysing change-from-baseline' tpmorris.substack.com

New post on dealing with missing baseline values in randomised trials that analyse change-from-baseline
open.substack.com/pub/tpmorris...

Advanced Confounding Adjustment (ACA). Dates June 16-20, 2025. Instructors: Joy Shi, Barbra Dickerman, Miguel Hernán.

Interested in using g-methods with time-varying confounders? 💡

Advanced Confounding Adjustment (ACA) teaches inverse probability weighting, parametric g-formula, & more.

📆 June 16-20, 2025

Taught by Joy Shi, Barbra Dickerman, @miguelhernan.org.

Register now:
causalab.hsph.harvard.edu/courses/

Accounting for Missing Data in Public Health Research Using a Synthesis of Statistical and Mathematical Models Introduction: Missing data is a challenge to medical research. Accounting for missing data by imputing or weighting conditional on covariates relies on the variable with missingness being observed at ...

It's a weird time to post about my research given ongoing events, but I'm going to share a new preprint

It's my 3rd paper in a series on synthesizing statistical and mathematical models, oriented to be more of an introduction with a NHANES example

arxiv.org/abs/2503.02789

addendum really gets into at all.

That's a quote from the ICH E9 addendum I think there. I think this document has population estimands in mind rather than sample ones. But as others have rightly pointed out to me, trial participants are definitely not random samples from well defined populations, and this isn't something the...

What is meant by a 'while on treatment' estimand? thestatsgeek.com/2025/03/03/w...

Want to learn the fundamental principles of #ClinicalTrials?🔬

Explore key issues in design, conduct, analysis & reporting on our hybrid short course. Designed for clinical research professionals, managers & scientists.

Runs 7 - 11 July 2025

Apply now✍️ bit.ly/2NldNE4

A Comparison of Statistical Methods for Time‐To‐Event Analyses in Randomized Controlled Trials Under Non‐Proportional Hazards While well-established methods for time-to-event data are available when the proportional hazards assumption holds, there is no consensus on the best inferential approach under non-proportional hazar...

doi.org/10.1002/sim....

Our paper "A Comparison of Statistical Methods for Time-To-Event Analyses in Randomized Controlled Trials Under Non-Proportional Hazards" got published today 🎉

We describe commonly used methods, and compare their performance in a simulation study across different scenarios.