James Steele
LightNews LightNews
banner
jamessteeleii.bsky.social
James Steele
@jamessteeleii.bsky.social
780 followers 530 following 5.1K posts
I'm just a guy who's an academic for fun
Posts Replies Media Videos
jamessteeleii.bsky.social
Day 16... Back to the long ones!

Please keep the support for me and @catsprotection.bsky.social coming 🙏

www.justgiving.com/page/james-s...
February 16, 2026 at 8:47 PM
jamessteeleii.bsky.social
Day 15... Apple Health definitely shafted me today... Morning walk alone was 7.3 miles when I mapped it out (from Trinity College to Papal Cross and back). Been the most obvious discrepancy in distance covered Vs recorded (others have been maybe up to a mile but today I think I'm missing 3-4 miles).
February 15, 2026 at 8:36 PM
jamessteeleii.bsky.social
Student
February 15, 2026 at 12:54 PM
jamessteeleii.bsky.social
Day 14 down... We just tipped over the halfway point to 500 miles today bang on time. At a total of 252.3 miles now.

Donate here for the kitties: www.justgiving.com/page/james-s...

@catsprotection.bsky.social
February 14, 2026 at 9:30 PM
jamessteeleii.bsky.social
So even if the effect is constant across the broader population, the sample SMD ends up large because of the selection process unless corrected.
February 13, 2026 at 7:54 PM
jamessteeleii.bsky.social
POV: some exercise science bro is lecturing you on philosophy of science
A presenter stands at a lectern during the Perspectives on Scientific Error conference, pointing to a projected slide titled “Precise Deductions” that shows hypothesis ranges and function-form diagrams.
February 13, 2026 at 5:01 PM
jamessteeleii.bsky.social
Day 12 down... Found another friendly kitty getting in some extra miles over lunch break. Bonus that my talk went down well today and inspired some wonderful discussion with attendees #PSE8

www.justgiving.com/page/james-s...
February 12, 2026 at 9:59 PM
jamessteeleii.bsky.social
Wow, we just tipped over the £1500 mark!!!

Thanks everyone for your generosity... Let's see if we can get to £2000!!!

Donate here: www.justgiving.com/page/james-s...
February 11, 2026 at 9:29 PM
jamessteeleii.bsky.social
Day 11... First day of the Perspectives on Scientific Error Conference so short distance covered today. No stress though as I'll be catching back up on these few short days once back.
February 11, 2026 at 9:13 PM
jamessteeleii.bsky.social
Day 10 down... Crossed 200 miles today with the longest one so far. Trying to get some extra miles to make up for lower mileage next few days whilst at #PSE8 conference in NL.

Don't forget to donate and support!

www.justgiving.com/page/james-s...
February 10, 2026 at 9:32 PM
jamessteeleii.bsky.social
Day 8 down... Missed posting a couple days here but 167.9 miles covered so far!

Keep donations coming to support the kitties 😻

www.justgiving.com/page/james-s...
Apple health screenshot showing 44772 steps and 23.8 miles walked
February 8, 2026 at 6:27 PM
jamessteeleii.bsky.social
You know whose name hasn't come up in any of the Epstein files?

Himmel the Hero
February 8, 2026 at 5:50 PM
jamessteeleii.bsky.social
Day 5... And that's 102.2 miles covered now for @catsprotection.bsky.social!

Long slow plodding today with a shin injury (after clumsily nailing it on the footstool) but managed to get there.
Screenshot of Apple health showing 41496 steps and 18.9 miles walked.
February 5, 2026 at 7:53 PM
jamessteeleii.bsky.social
We're about 10% of the way through extraction building the largest meta-analytic database of resting energy expenditure data as part of a big evidence synthesis project here at MacroFactor!
Forest plot showing study- or group-level mean resting energy expenditure (REE) in kcal/day with corresponding standard errors. Each horizontal line represents the SE around a mean estimate (shown as a point), with estimates ordered from lowest to highest REE along the vertical axis. Mean REE values range approximately from 900 to 2,600 kcal/day. The distribution forms an S-shaped pattern, with most estimates clustered between about 1,300 and 1,800 kcal/day and greater variability at higher REE values.
February 5, 2026 at 11:37 AM
jamessteeleii.bsky.social
Day 4... Still roaming.

This morning was a struggle with a bit of an injury... More donations and support would be very much welcomed 😸

www.justgiving.com/page/james-s...

@catsprotection.bsky.social
A screenshot of Apple health showing 43027 steps and 23.3 miles walked
February 5, 2026 at 10:13 AM
jamessteeleii.bsky.social
Day 3... Still walking to support @catsprotection.bsky.social

You support the kitties and I by donating here and sharing the link too

www.justgiving.com/page/james-s...
February 4, 2026 at 10:22 AM
jamessteeleii.bsky.social
Day 2... And yesterday we hit the £1000 mark for donations. Let's keep it going and see how high we can get.

Please share far and wide www.justgiving.com/page/james-s...
A screenshot of Apple health showing 35473 steps and 19.8 miles walked. A screenshot of the linked just giving page showing £1000 reached in donations.
February 3, 2026 at 9:43 AM
jamessteeleii.bsky.social
Day 1 down... Keep the donations coming! Every little bit helps Cats Protection do so much great work.

www.justgiving.com/page/james-s...
February 1, 2026 at 7:33 PM
jamessteeleii.bsky.social
Oooh we got some DAGs coming along to visualise our simulations - needs some tweaking but it's getting there.
January 28, 2026 at 4:44 PM
jamessteeleii.bsky.social
A man sits at a small folding table outdoors on a university campus, holding a mug and smiling, in the style of the “Change My Mind” meme. A sign taped to the front of the table reads “CHANGE MY MIND.” Another sign on the tabletop reads: “Most quant vs qual methods dichotomies as presented are fallacious and more to do with differences in ontological and epistemological differences of the researchers themselves.” Trees, benches, and campus buildings are visible in the background.
January 19, 2026 at 11:33 AM
jamessteeleii.bsky.social
As if I haven't banged on about enough by now... looking forward to continuing to talk about how developing and trying to test strong theories is a pretty damned useful way of going about doing science.

Here's my abstract for the 8th Perspectives on Scientific Error Workshop in Leiden, NL.

#PSE8
The image shows the abstract for my talk "The value of strong theory in intervention research: an example from the field of exercise science" at the upcoming 8th Perspectives on Scientific Error Workshop - you can find it in the program here: https://docs.google.com/document/d/1rt9ToVs1EkEuTbWod4st2b6StUQr5IhyTP1jhGskr4E/edit?usp=sharing
January 13, 2026 at 4:30 PM
jamessteeleii.bsky.social
Sigh... note to self, don't do stuff quickly in between trying to do your day job 😅

Now(!) it's the correct variance structure for intercepts (1 | survey/effect).
January 12, 2026 at 1:39 PM
jamessteeleii.bsky.social
Corrected errors on variance calculations and structure (thx @jbakcoleman.bsky.social) and added corresponding interval estimates.

Note, see Joe's comments too... good points to consider. My aim wasn't really anything more than just pointing out you can meta-analyse these things if you want.
Multi-panel horizontal bar chart showing self-admitted rates (percent on x-axis) of engaging in various questionable research practices (QRPs) at least once. Each panel represents a distinct QRP, including excluding data based on results, failing to report all conditions, falsifying data, HARKing, hiding demographic moderators, opportunistic p-value rounding, optional stopping, selective outcome reporting, selectively reporting performed analyses, selectively reporting what “worked,” selectively including covariates, and selectively switching analyses. Colored horizontal bars indicate estimates from individual studies. In each panel, a vertical red line marks the meta-analytic estimate across studies, and a surrounding red shaded band represents the 95% confidence interval for that estimate. Prevalence varies widely across practices and studies, with higher rates for selective reporting behaviors and HARKing, and consistently low rates for falsifying data. Alt text generated by chatgpt.
January 12, 2026 at 1:23 PM
jamessteeleii.bsky.social
Ah, tbf I assumed it was because of the inverse variance weights of the larger surveys in the sample. Prompted me to check and realised I'd copy-pasted wrong snippet of code for variance calculations. Also, in haste forgot survey level random effect 🙃 Here we go... thanks!
January 12, 2026 at 1:13 PM
jamessteeleii.bsky.social
A quick meta-analytic* estimate of QRPs across surveys added to @lakens.bsky.social collation of results from them (lakens.github.io/statistical_... pull request made).

Red line is the point estimate (intervals are tiny so not plotted)

*meta-analysis of proportions using the logit transformation
Multi-panel bar chart showing self-admitted rates (percentages on the x-axis) of engaging in various questionable research practices (QRPs) at least once, based on multiple published studies (color-coded by study). Each panel corresponds to a specific QRP, including: excluding data based on impact on results, failing to report all conditions, falsifying data, HARKing, hiding demographic moderators, opportunistically rounding p-values, optional stopping, selectively reporting outcomes, selectively reporting performed analyses, selectively reporting what “worked,” selectively including covariates, and switching analyses selectively. Within each panel, horizontal bars represent individual study estimates of the percentage of respondents admitting to that practice. A vertical red line in each panel indicates the meta-analytic estimate across studies for that QRP. Overall, reported prevalence varies widely by practice and study, with some QRPs (e.g., selective reporting and HARKing) showing higher admission rates than others (e.g., falsifying data).
January 12, 2026 at 9:57 AM