Ben Harrap
LightNews LightNews
banner
bharrap.bsky.social
Ben Harrap
@bharrap.bsky.social
2.2K followers 750 following 870 posts
doer of data things at Yardhura Walani
vice president of statsoc.org.au
I do #rstats, surveys, biostatistics, #IDSov
on Ngunnawal and Ngambri lands
not Indigenous, non-binary (they/them)
benharrap.com
Posts Replies Media Videos
bharrap.bsky.social
Maybe in a few promotions' time 🥲
Two panel meme of a excited then disappointed guy. Excited panel reads "the license plate YMDHMS is available" Disappointed panel reads "it's $3600"
November 10, 2025 at 10:03 AM
bharrap.bsky.social
I've been wearing band shirts for years but today is the first time someone's thrown up the evil eye/devil horns

Rock on Bunnings dude 🤘
A black t-shirt with the word 'emperor' and a stylised image of a death wielding a scythe atop a horse
November 6, 2025 at 2:26 AM
bharrap.bsky.social
I've made a new RStudio theme based on these nice logs. Enjoy #rstats!

benharrap.com/post/2025-10...
Logs cut from a gum tree. They have silvery bark and wood with an orangey hue A preview of R code with a colour palette based on gum logs
October 31, 2025 at 12:02 AM
bharrap.bsky.social
TBH the only use of this domain name should be to display images of pet turtles wielding sai
raphael, one of the teenage mutant ninja turtles, wielding two sai
October 12, 2025 at 11:58 PM
bharrap.bsky.social
Great pointer from @nrennie.bsky.social on making data visualisations

We as makers spend hours with the data and creating the visualisation, but we need to consider those who are looking for 5 seconds

#WOMBAT2025 #dataviz
A slide about accurate visualisations asking Is the chart appropriate Are the axis ranges reasonable Is it intuitive Is the 5 second conclusion the same as the 5 minute conclusion
September 29, 2025 at 11:11 PM
bharrap.bsky.social
In my most spectacular kitchen fail to date, I forgot I undid the lid on the passata and proceeded give it a big shake...

I couldn't help but laugh and show my toddler the sheer mess I just made

🙈🙈🙈
A kitchen benchtop and chopping board splattered with passata
September 27, 2025 at 9:30 AM
bharrap.bsky.social
My favourite bush, pink diosma
An Australian native bush in bloom with cute little pink flowers
September 25, 2025 at 8:45 PM
bharrap.bsky.social
Time for the SSA ACT branch monthly(ish) talk on tobacco control

Grace has many publications on Tobacco control - check out her ANU profile:

researchportalplus.anu.edu.au/en/persons/g...

#statssky #episky
The title slide from a presentation. It reads "smoking guns - generating evidence to inform tobacco control. Associate professor Grace Joshy"
August 26, 2025 at 7:55 AM
bharrap.bsky.social
Gotta christen the new house with maybe too much pancetta @elyobo.bsky.social
Diced pancetta slowly rendering in a pot
August 17, 2025 at 7:08 AM
bharrap.bsky.social
lmao would recommend
A response from the Claude LLM after being instructed to respond like an aggressive New Yorker: Alright, alright, you want a function that strips ALL the whitespace, including those sneaky unicode characters? No problem, I got you covered. Here ya go - this'll handle everything from your basic spaces to those weird unicode whitespace characters that like to hide in your data: There ya have it! I gave you THREE versions because I'm thorough like that: Basic version - Uses \\s+ which catches 99% of whitespace including unicode. The perl = TRUE makes sure it handles unicode properly. Stringr version - Same thing but uses stringr if you're one of those tidyverse people. Comprehensive version - This bad boy explicitly targets specific unicode whitespace characters that sometimes slip through. I'm talking non-breaking spaces, em spaces, figure spaces - the whole gang. The \\s+ pattern is your friend here - it matches ANY whitespace character (spaces, tabs, newlines, unicode whitespace) and the + means it'll grab consecutive whitespace characters all at once. Just paste whichever version works for your situation and boom - no more whitespace problems. You need anything else or we good here?
July 31, 2025 at 10:58 PM
bharrap.bsky.social
Sure, see the video

I've realised that it isn't the curly braces alone that cause the anonymous function to appear in the bookmarks - it only appears when the .fns argument is named (same for `.f =` in `map` and `.fn = ` in `rename_with`)
July 27, 2025 at 10:53 PM
bharrap.bsky.social
Accidental #rstats / RStudio discovery of the day:

You can auto-complete folder locations 🤯

It seems obvious now, but I'd always assumed being inside quotes would've meant this wouldn't work!
July 21, 2025 at 12:30 AM
bharrap.bsky.social
Folks this is a very much joke

And yes I need to say it, some of y'all been googling 👀

#rstats #databs
A graph from google trends showing a spike in search activity for "kitchensink r package"
July 17, 2025 at 6:45 AM
bharrap.bsky.social
"my favourite generative AI model is my brain" coming to a radio station near you

#genAI #databs
July 14, 2025 at 11:41 AM
bharrap.bsky.social
Looks pretty good! Just got to strain the fat and make some gravy
July 6, 2025 at 7:04 AM
bharrap.bsky.social
See you in six hours
A joint of lamb leg sitting atop onions, carrots, and potatoes in a roasting tray
July 6, 2025 at 12:15 AM
bharrap.bsky.social
Absolutely cracking Yorkshire puds tonight
Twelve crispy, golden-brown Yorkshire puddings
June 29, 2025 at 7:46 AM
bharrap.bsky.social
More like Cohen's DON'T #statssky
Shrek ew no yes please meme Top panel: Shrek ew no face and the text "Sample size calculation based on Cohen's d" Bottom panel: Shrek yes please face and the text "Sample size calculation based on the outcome's original scale"
June 24, 2025 at 12:16 AM
bharrap.bsky.social
I thought getting an espresso machine would be a one-off expense but I have since discovered the wonderful world of accessories and upgrades

@elyobo.bsky.social I think you will appreciate
A sky-blue profitec GO espresso machine next to a coffee grinder on a kitchen bench
June 13, 2025 at 2:23 AM
bharrap.bsky.social
I had some fun coming up with a simple function written in base R, which randomly picks a pronoun for you. I wrote a lil' blog post about it too.

Unfortunately, English is annoying (my function is perfect)

#rstats #quarto

benharrap.com/post/2025-06...
A snippet of code which defines a list of pronouns and a function that randomly selects a pronoun from said list.
June 12, 2025 at 6:43 AM
bharrap.bsky.social
The number one thing I want people to take away is:

None of the avenues toward reducing rates of PPHs require children to be removed from their families

In fact, providing culturally safe primary healthcare and social supports BEFORE removing children would likely lower rates of removal

12/
A screenshot of text from the article's discussion: When considering the mechanisms for prevention suggested by Anderson et al., most conditions were preventable through early access to primary care. Only dental conditions, gastroenteritis, and skin infections were considered preventable through health promotion. However, these conditions constituted 30%–50 % of conditions across the three age brackets, suggesting health promotion can play an important role in policy design to reduce rates of these PPHs. Underlying this, however, is our finding that the rate of PPHs increased with increasing socio-economic disadvantage. If the household environment and access to timely primary healthcare are dependent on financial capacity, such as paying for housing with more bedrooms to avoid overcrowding or visiting general practitioners and dentists that charge gap fees to access them earlier, the influence of health promotion will be limited for the most socio-economically disadvantaged families while financial barriers exist. Screening for healthcare needs, referrals to specialists, changing household environment and no longer being maltreated could all contribute to lower rates of PPHs. What is important to realise is that these preventive measures can all occur without having to remove a child from their family—through early intervention and support services for families.
June 3, 2025 at 4:36 AM
bharrap.bsky.social
I ran models for specific categories of conditions, but the estimates became much more uncertain

Interestingly, PPHs for dental conditions and otitis media *increased* following first placement in OOHC. Possibly a result of increased surveillance via health screening upon entry?

10/
Six different plots of point-estimates with 95% confidence intervals, with rates of PPHs on the Y-axis and age from 1 to 15 on the X-axis. Plot 1 is for skin infections and shows a slow decrease in PPH rates with increasing age, with a reasonable reduction following first OOHC placement Plot 2 is for respiratory conditions, showing rates spike in the first two years of life then reduce dramatically to be almost zero. Confidence intervals overlap considerably for the estimates pre/post first-placement in OOHC Plot 3 is for gastroenteritis and has the same pattern as plot 2 Plot 4 is for dental conditions, showing barely any admissions until age 3, then rising to its peak around ages 5-6, then reducing through to age 15. In this plot, rates increase following first OOHC placement Plot 5 is for otitis media and also shows an increase in rates following first OOHC placement. Rates peak at age 2 and reduce each year through to age 15 Plot 6 is for pneumonia and is difficult to deduce anything from. Rates are highest in the first two years of life but confidence intervals are very wide.
June 3, 2025 at 4:36 AM
bharrap.bsky.social
The modelling suggested that placement in OOHC plausibly reduced rates of PPHs from anywhere between 3% to 18%

Consistent with PPH literature in Australia, rates of PPHs increased with increasing remoteness and socioeconomic disadvantage

9/
Plots of the point-estimates and confidence intervals for rates of PPHs. Rates are on the Y-axis and age from 1 to 15 is on the X-axis. Each plot is for a different remoteness category - all areas, major cities, inner regional, outer regional, remote, and very remote. For all plots, rates of PPHs peak in the first two years of life, then reduce until around age 5 where they remain consistently low. As remoteness increases, so too does the rate of PPHs in the first few years of life.
June 3, 2025 at 4:36 AM
bharrap.bsky.social
Aim 3 attempts to estimate the effect of first OOHC placement on rates of PPHs using mixed-effects negative binomial regression (via {glmmTMB} and {emmeans})

Model included an exposure*prepost interaction, along with age, remoteness, and socioeconomic decile in formats determined using {mfp2}

8/
A screenshot of the article text: Our modelling of the effect of first placement in OOHC on the rate of bed days for PPHs for ever-placed children provided an estimated incidence rate ratio of 0.89 (95% CI: 0.82 to 0.97), suggesting that on average the rate of PPH bed days reduced by 11% (95% CI: 3% to 18%) in the period following a child’s first placement in OOHC, with a reduction ranging from 3% to 18% also being plausible. Examining the marginal rate of PPH bed days per 1,000 days at risk for children aged 1 to 14 and across remoteness areas (Figure 1), found that the estimated difference between before and after first placement in OOHC was greatest in the first year of life. For all children, increasing age was the largest driver for reductions in the rate of PPH bed days, with the rate falling most rapidly from ages one to five. Increasing remoteness was associated with an increase in the rate of bed days for never- and ever-placed children.
June 3, 2025 at 4:36 AM
bharrap.bsky.social
Aim 2 compares the conditions identified in PPH admissions, as they are defined by a specific set of conditions which are considered preventable

Across age groups (0-4, 5-9, 10-14), the proportion of admissions with a given condition were basically the same for ever- and never-placed groups

7/
A screenshot of a table from the article, showing the top ten conditions seen in PPH admissions across ages 0-4, 5-9, and 10-14
June 3, 2025 at 4:36 AM