Sam Abbott

I imagine a very similar idea can be used for wastewater surviellance models

Oo this is fun!

Unfortunately due to speaker illness this seminar has had to be cancelled.

In March 2024 I was reading R docs and learned 🧵

1. doc for NaN and is.nan() says: "‘NaN’ means ‘Not a Number’" and "NA, ‘_Not Available_’ which is not a number as well"

-> so neither is "a number".

Yes agree and/or the right people have both the tools and the data they need.

#epinowcast seminar tomorrow at 3pm UK time/10am US East time: David Hodgson talking about Serological data and the seroanalytics suite of tools

www.epinowcast.org/seminars/202...

This is despite a lot of work and focus in this area over the last decade which I think flags that the incentives and approaches we are using as levers are flawed.

I still find it profoundly shocking we don't have the tools and methods we need to estimate many of these delays in real time at the kind of granularity needed in order to inform most measures robustly.

This is a nice piece and makes some good points. Something it mentions is that work is being done to make the tools we need but ...

H3N2 preprint: there are concerns of a severe incoming influenza season due to the drifted H3N2 K clade. We at @psioxford.bsky.social analysed epi data and ran scenario models to see what we could discern about K clade transmission dynamics: zenodo.org/records/1770....

(1/18)

Model is: b3 <- scasm( y ~ s(x0, bs = "bs", k= k) + s(x1, bs = "sc", xt = "m+", k = k) + s(x2, bs = "bs", k = k) + s(x3, bs = "bs", k = k), family=poisson, bs=200 ) The second smooth `s(x1) is a shape constrained smooth with a positive monotonicity constraint (xt = "m+"). The `bs = 200` arguments uses 200 boostrap samples, which generates bootstrap distributions for each coefficient in the model. These bootstrap samples respect the shape constraints, while the usual +/- 2 SE credible intervals may not. The uncertainty in the partial effects is shown by two credible interval bands; a dark blue central band is a 68% Bayesian credible interval, while the lighter blue outer interval is a 95% Bayesian credible interval. The background of each panel is light grey with white grid lines, in a similar style to ggplot2's default theme.

A new release of the mgcv #RStats 📦 is out on CRAN and Simon Wood (U Edinburgh) has added some significant new features despite the small bump in version number:

🌟 scasm() for estimating GAMs with shape constrained smooths. Can be used with any family & smoothness selection is via the EFS method

For some reason it is much more of a pleasure when it is some software or small idea than a whole paper. Maybe because we have less formal ways to refer to those?

which evaluates and builds on some of the work in hashpring
(which really was totally down to Carl than me).

github.com/epinowcast/h...

Methods for Reproducible Comparison of Strategies in Stochastic Modelling Stochastic simulations are often used to model real-world phenomena such as infectious disease dynamics. In this modelling, differing strategies are often compared to one another by comparing the mode...

Always a pleasure when you see a package or some scratch work being used as part of a longer form piece of work.

Nice paper here by Rob Sunnucks et al. www.medrxiv.org/content/10.1...

Ten Simple Rules for AI-Assisted Coding in Science While AI coding tools have demonstrated potential to accelerate software development, their use in scientific computing raises critical questions about code quality and scientific validity. In this pa...

Ten Simple Rules for AI-Assisted Coding in Science arxiv.org/abs/2510.22254 - our latest, led by @ericwbridgeford.bsky.social

Really nice to see this being used - a real blast from the past. If anyone is keen to help get this back on CRAN or do other work on it that would be amazing!

Logo for the #TidyTuesday Project. The words TidyTuesday, A weekly data project from the Data Science Learning Community (dslc.io) overlaying a black paint splash.

TidyTuesday is a weekly social data project. All are welcome to participate! Please remember to share the code used to generate your results! TidyTuesday is organized by the Data Science Learning Community. Join our Slack for free online help with R and other data-related topics, or to participate in a data-related book club! How to Participate Data is posted to social media every Monday morning. Follow the instructions in the new post for how to download the data. Explore the data, watching out for interesting relationships. We would like to emphasize that you should not draw conclusions about causation in the data. Create a visualization, a model, a shiny app, or some other piece of data-science-related output, using R or another programming language. Share your output and the code used to generate it on social media with the #TidyTuesday hashtag.

World map showing estimated TB incidence rates per 100,000 population. Countries range from light green (low incidence) to dark purple (high incidence), with highest rates in Sub-Saharan Africa and South/Southeast Asia. Map sourced from WHO and rendered using getTBinR::map_tb_burden() in R.

@dslc.io welcomes you to week 45 of #TidyTuesday! We're exploring WHO TB Burden Data: Incidence, Mortality, and Population!

📁 https://tidytues.day/2025/2025-11-11
📰 https://samabbott.co.uk/getTBinR/index.html

#RStats #PyData #JuliaLang #DataViz #tidyverse #r4ds

Liked this: "In conclusion, thinking about modelling as experimentation does not cage creativity; it structures it, giving us better tools to ask sharper questions, present clearer answers and build cumulative science."

I really enjoyed this - luckily found it just before having a call with John about workflows for IDM vs just after.

You're absolutely right! 😜

(I mean this all in a context of talking about code not day to day!)

For example, I really like communicating briefly and do so with LLMs as it works for them. However, with people its important to add qualifies etc.

Similarly, in a review the LLM can be told to change something small on bulk without a worry but for a person this might take a really long time.

I spend a lot of my day interacting with LLMs and reviewing LLM generated content (mostly all via Claude Code). I'd really like it to communicate in a way that is more clearly not human as I find sometimes that when I talk to actual people I might not have switched communication style.

5 minutes!

This is today at 3pm UK time

For this week’s #epinowcast seminar (as ever the first Wednesday of the month at 3pm UK time) we have Lina Cristancho Fajardo talking about:

Why heterogeneity in mosquito exposure is necessary to forecast chikungunya outbreaks in Reunion Island.

www.epinowcast.org/seminars/202...

Now I'm also looking for a research software engineer to implement a pile of research results to R packages loo, posterior, bayesplot, projpred, priorsense, brms or/and Python packages ArviZ, Bambi and Kulprit. Apply by email with no specific deadline (see contact info at users.aalto.fi/~ave/)

Untargeted longitudinal ultra deep metagenomic sequencing of wastewater provides a comprehensive readout of expected and unexpected viral pathogens Wastewater surveillance has become a powerful tool to monitor circulating viruses at a community level. Currently, most wastewater surveillance efforts use target-based approaches such as quantitative...

Can you take a quarter cup of composite sewage, simply ask ‘what’s in there?’, and find out all of the pathogens circulating in that community?

That is the question we asked in our latest pre-print.

Turns out you can.
1/
www.medrxiv.org/content/10.1...

I just had a look for tools etc to do some data analysis on this for an individual or research more generally but in the quick skim I had time for didn't find anything.

Does anyone know of good resources?

Had another similar conversation about software papers being a bit pointless and just let people cite the software.