Thread by @elizsn

Our meta-analysis is published - prek/kindergarten RAN *robustly* relates to later reading ability in English.
Please share with those 🧑‍🏫👪who need to advocate for including #RAN in a #dyslexia #screening battery. Open access paper! #RANmeta
ila.onlinelibrary.wiley.com/doi/full/10.10…
Thread:

Dr. Sean McWeeny started this meta-analysis in 2018. He has become such a star in this area of work. He co-developed the SnowGlobe tool to automate searching each paper's refs and future citations to ensure we had excellent coverage of the literature. snowglobe.soc.northwestern.edu

Sean, PhD student Jinnie Choi, & a great team of undergraduates in the @nulearnlab screened well over 4,600 articles for inclusion in the analyses. Final count included 67 studies with 10,513 participants. We used analyses that let us look at multiple effect sizes within studies.

We measured the overall relation between
early RAN ➡️ later reading at r = -.38
(faster RAN time = better reading). This fits with findings from previous concurrent meta-analyses and one smaller longitudinal one that focused on reading comprehension across languages.

Here's the cool part 😎- we focused on testing moderators of the RAN-reading relation. This means, we looked at what factors can alter the strength of the RAN-reading relation. I have been curious about these things for SO long!

We analyzed lots of potential moderators about the RAN task itself: number of total items, number of unique items, type of items, whether it was published or not.
About the sample: their age/grade at screening, and proportion of kids with dyslexia risk.

Also about reading outcomes 📖, type of reading measure (timed/untimed, real/nonword, fluency, comprehension).
Our models show that RAN is extremely consistent in its relation with reading, almost none of these things significantly changed the RAN-reading relation. 📊

One thing that mattered: alphanumeric (letter🔠or digit naming🔢) was more strongly related to later reading than object🐶⭐️🖐️or color 🟨🟥naming (Δr = 0.13; p=.01). This was not driven by age at assessment (eg, older kids are more likely to be given an alphanumeric task)...

...it means that alphanumeric RAN is just more closely related with reading. This supports a shared cognitive resource model, or RAN as a "microcosm" of reading as @MaryanneWolf_ and I noted 10 years ago. annualreviews.org/doi/abs/10.114…

Also cool: a reviewer suggested that we run semi-partial correlations controlling for phonological awareness (PA) because so many studies test both (n=5,452). We did, and the result of rsp = −.25 shows that RAN makes a significant, unique contribution to later reading beyond PA.

Though we found that RAN better predicted real word than nonword reading without controlling for PA, this was nonsignificant when controlling for PA. When we controlled for PA, we also saw that RAN was significantly better in predicting reading efficiency > accuracy. Makes sense!

Takeaways for screening 🏫in schools! 👉ASSESS RAN. It's quick and easy. Anytime before 1st grade works, but there is no reason to wait. You can make your own RAN task, but you might as well use a standardized one so you have a representative comparison group.

Previous review paper with more info for educators: learnlab.northwestern.edu/wp-content/upl…
Thanks everyone for reading! Great work Sean 👨‍🎓! Go RAN!