Stats For Bios
LightNews LightNews
banner
statsforbios.bsky.social
Stats For Bios
@statsforbios.bsky.social
2K followers 420 following 1K posts
Statistics for biologists, medics, statisticians. Anyone really. By Tim Lucas, lecturer in epi/stats.
Posts Replies Media Videos
statsforbios.bsky.social
Sorry this has all got me on a quick nostalgia kick. This is the venue I learned to love live music in (the Freebutt). It got closed down because they built residential houses next to it, people moved in and didn't like the noise. And now it's just nothing. What an absolute waste.
A run-down, closed-up venue. Looks pub size.
November 20, 2025 at 3:38 PM
statsforbios.bsky.social
Anyone know this film? It's on in the bar and appears to be completely bonkers. But I have no idea what it is
November 9, 2025 at 5:26 AM
statsforbios.bsky.social
If anyone's going to ASTMH a day early and likes punk, pretty fun looking gig on Saturday. I'll be there if my flights behave.
files.elfsightcdn.com/eafe4a4d-343...
Gig poster. Bovine sex club venu.e Decade, wolfcharge absolutely and manic. 9pm, nov 8th, $10.
November 7, 2025 at 9:18 PM
statsforbios.bsky.social
New paper, great work by Hannah Worboys. Feasibility study of the P-STEP app. "At 12 weeks, the mean (sd) System Usability Score was 61.68 (22.9)". And lots of useful prelim data for larger studies. #EpiSky #EnvironmentalEpi #copd #asthma
Four histograms. a) Usage per week (mostly between 1 and 6 times a week), b) likely to recommend (mostly likely), c) app rating (mostly 4-5 stars) and d) attitudes to future use (almost 50/50 on future use).
October 20, 2025 at 9:33 AM
statsforbios.bsky.social
Wait, stata costs twice as much if you want to run it using 2 cores? And even more for 4 cores? Lol.
A table of stata prices. Standard is $94. "MP 2 core" is $275.
September 29, 2025 at 4:06 PM
statsforbios.bsky.social
Patent infringment from Wallace (& Gromit) coming any day now.
July 25, 2025 at 8:48 AM
statsforbios.bsky.social
Dawww. A baby bee.

Please don't a) tell me it's not a baby bee or b) tell me it's some horrible invasive. Enough of a rollercoaster day already.
July 23, 2025 at 1:22 PM
statsforbios.bsky.social
Got case data aggregated to admin areas?
Got important covariates in high-resolution rasters/grids?

We are running a free online training workshop on user-friendly, interactive software for this use-case. Disaggregation regression via the shinyapp disagapp.

github.com/simon-smart8... #malaria
Screenshot of disagapp. Shows malaria data from madagascar aggregated to admin regions on an interactive map. A nicely designed user interface provides modelling capabilities.
July 17, 2025 at 2:32 PM
statsforbios.bsky.social
You often have to choose the number of clusters. If clusters are clearly arbitrary and just a modelling tool, fine, look at some plots and choose.

But when the main result of the paper is "there are 4 clusters and this reflects reality and here's our description of the clusters". It's so weak.
6 plots showing metrics of a model with different numbers of components, 1-10. Log likelihood, AIC, BIC and others are shown. They all very smoothly improve with number of clusters. No clear cut jumps or elbows.
July 11, 2025 at 9:03 AM
statsforbios.bsky.social
I so often find dual axis graphs confusing while I sort out what's what. Having it spelled out like this makes a huge difference. #DataViz
A dual axis graph. Two lines showing two variables changing through time. Text, next to each line, says what the variable is and which y axis it corresponds to.
July 2, 2025 at 9:05 AM
statsforbios.bsky.social
Our new paper led by Jenicca Poongavanan. Predicting #dengue in Africa using disaggregation regression (bottom right in the inset image) and comparing to other methods. www.biorxiv.org/content/10.1...
A compound figure. The main image is of Africa with colour showing the number of models that agree in each pixel. 4 models agree in areas of west africa. 3 models agree in a large area should of the Sahara. The inset has four binary prediction maps for the different models.
June 23, 2025 at 9:21 AM
statsforbios.bsky.social
New paper led by @esnightingale.bsky.social. Delay to diagnosis is important for visceral leishmaniasis control. Longer delay -> more transmission.

Can we understand these delays, factors correlated with them and spatial distribution. bmcglobalpublichealth.biomedcentral.com/articles/10....
3 maps of Bihar India. Map of estimated delay is diagnosis. Posterior probability that the delay is greater than 30 days. And the raw block mean delay.
June 11, 2025 at 3:18 PM
statsforbios.bsky.social
Messing around with my last.fm scrobble data again. Each year, who were my 5 favourite new (to me) bands? #RStats
A screen shot of a table from R. Columns are artist, scrobbles and year. Excerpt: 87 Andy B & The World 282 2022 88 Celeste 158 2022 89 Israel Vibration 156 2022 90 Astronoid 131 2022
June 7, 2025 at 4:10 PM
statsforbios.bsky.social
May 21, 2025 at 1:05 PM
statsforbios.bsky.social
Does America have anything roughly equivalent to a potato scone? I can imagine they'd be very popular.
en.wikipedia.org/wiki/Tattie_...
A potato scone. It is a fried pancake type thing. Golden brown, glistening slightly from the butter, served with a mega crunchie looking thing that I can't remember of the name of and a fried tomato. The adjacent fried bacon and black pudding is implied but not directly shown in the image. Availability of brown sauce is open to interpretation.
May 10, 2025 at 7:13 PM
statsforbios.bsky.social
New work by my student Flo Goemans (and others).
No change between sexes in accuracy of polygenic risk scores for idiopathic pulmonary fibrosis. Surprising given higher prevalence in males. publications.ersnet.org/content/erjo...
AUC curves for prediction of idiopathic pulmonary fibrosis using polygenic risk scores. Separate curves for male, female and overall but all three curves are very similar.
May 9, 2025 at 9:48 AM
statsforbios.bsky.social
If I'm reading this right, it's dual axis, on totally different scales (% Vs millions) but the numbers just happen to line up exactly. Pretty crazy and more confusing than if they were different. #DataViz
Horizontal bar plot of ww2 deaths by country. Has deaths in millions and deaths as a percent of the country. Top and bottom axis have identical numbers on the axis ticks.
April 30, 2025 at 6:36 PM
statsforbios.bsky.social
Image from the film quadraphenia. The main character (whose name I've forgotten), stands in an alley in typical mod clothes.
April 16, 2025 at 7:30 PM
statsforbios.bsky.social
Thanks #EuroCIM2025! Great conference. Now coding up my first ipw for malaria inference on my way home. #StatsSky #CausalInference #CasualInference
A photo of a laptop with maps of environmental variables in Madagascar and some R code. In a bar (in Brussels) with quite a nice plant.
April 11, 2025 at 1:17 PM
statsforbios.bsky.social
Could be worse.
Technical analysis (for trading) chart.
March 1, 2025 at 7:08 PM
statsforbios.bsky.social
Not sure I've ever used geom_smooth() like this to make essentially a calibration plot. Super easy.

Also, hot damn my simple model is well calibrated. Predictive performance is bad, but calibration is gooood. # Rstats
ggplot2 scatter plot with predicted probability of diabetes on x axis and diabetes (0 or 1) on y axis. Geom smooth creates a very straight line that almost exactly matches the reference slope=1, intercept = 0 line.
February 26, 2025 at 10:30 PM
statsforbios.bsky.social
Inelegantly answering my 9yr old niece's homework. #RStats
Photo of some R code. Nested loops.
January 22, 2025 at 7:34 PM
statsforbios.bsky.social
Painting social with Leicester biostats
November 25, 2024 at 8:02 PM
statsforbios.bsky.social
Nice to see disaggregation regression used in more areas.

"Our findings show that poorer communities are more prone to selling cattle in response to precipitation shortages, a practice that can erode long-term resilience and deepen inequalities."

www.researchsquare.com/article/rs-5...
Maps of cattle slaughter in South Africa. A) raw data aggregated to province level. B) Estimates of disaggregtted cattle slaughter at high resolution (16km pixels).
November 22, 2024 at 10:17 AM
statsforbios.bsky.social
Hooray! Our paper (led by Emily Nightingale) is out.

Inferring the regional distribution of Visceral Leishmaniasis incidence from data at different spatial scales.

w/ @grahammedley.bsky.social and others.

www.nature.com/articles/s43...
Panel plot. a) shows estimated coefficients from models with and without a spatial random field. NDVI is positive for both. LST is strong negative until you add a field. b) predicted rate of VL. High in the north of Bihar, India. c) The fitted random field. High in North Bihar as well.
November 21, 2024 at 10:58 AM