Two issues I like to make use of when instructing stats are:
1) Journal of the American Medical Affiliation (JAMA) visible abstracts. I’ve blogged about them earlier than.
2) Helpful instruments to generate faux information units that mimic actual information, and use these faux information units to show. See: Richard Landers‘ and Andrew Luttrell‘s web sites.
So, I used to be delighted after I noticed this not too long ago posted visible summary about Ewing-Cobbs et al. (2026) analysis on utilizing a particular CBT program to scale back stress in kids following a traumatic bodily harm.
| https://jamanetwork.com/journals/jamapediatrics/fullarticle/2848163 |
I’ve a brand new instance of an unbiased t check for sophistication. Yay! And I train tons of future nurses/PAs, so it’s doubly relevant.
Nevertheless, the authors said that the information wasn’t instantly accessible. Additionally, as soon as it’s accessible, they (very moderately) need to observe their information sharing. Which means that even when I might get their information, I should not be sharing it on this weblog.
I made a decision to create a dataset that mimics the findings. Whereas I’ve different instruments to take action (see above), I made a decision to strive utilizing GenAI this time. I figured that if I had extra descriptive information, I might make a greater pretend dataset. I discovered the CIs (in yellow) for the information factors within the tables:
Armed with means, pattern sizes, and CIs, I wrote this immediate:
“Hello! I would love you to create an information set for me. It ought to include two variables, “ReSeT” and “Normal Care”. The imply for ReSet ought to be 10.5, with a 95% CI of 8.1-12.9, and n = 44. The imply for Normal Care ought to be 14.7 with a 95% CI between 12.2 and 17.2, and n = 42. For the information you generate, each information level ought to be an entire quantity with no decimals.”
After some fanagling (Copilot could not completely generate the information within the parameters I said, however we obtained actually shut), I generated the information units (Right here is the .txt model, and right here is similar information, in JASP format for my fellow JASP customers/instructors)
How I’ll use it at school: Typically, stats instructors construction their class in order that college students clear up a thriller by analyzing information. That is nice. However it’s also good to work backward: Present your college students the visible summary. Have them establish the IV, the DV, and the findings. THEN have them analyze information that, whereas not the true information, does mimic the primary findings. It’s kind of like giving them a street map. They know precisely the place they’re going, however they nonetheless want to investigate the information themselves. Evidently, after I use information units that I create, I ALWAYS inform my college students that they don’t seem to be working with the precise information, however with information that mimics the true findings.
