Kevin M. King @kevinmking.bsky.social · 20/06/20255) LCGAs also are very sensitive to lots of other things, like the time scale you study, what time intervals you select, or the scale of your data. psycnet.apa.org/record/2006-... psycnet.apa.org/fulltext/200...psycnet.apa.orgAPA PsycNet 1112
Kevin M. King @kevinmking.bsky.social · 20/06/20256) LCGAs never replicate across datasets or in the same dataset. They usually just produce the salsa pattern (Hi/med/low) or the cats cradle (Hi/low/increasing/decreasing). This has misled entire fields (see all of George Bonnano's work on resilience, for example). psycnet.apa.org/fulltext/201...psycnet.apa.orgAPA PsycNet 6294
Karla Holmboe @karlaholmboe.bsky.social · 20/06/2025Interested in what you recommend as a better alternative for longitudinal data? 100
Andrew Percy @prof-andy.bsky.social · 20/06/2025The cats cradle paper (Ken Sher) is one of the greatest methodological publications ever. Crushed so many of my misconceptions of LCA. 061
Kevin M. King @kevinmking.bsky.social · 20/06/2025Latent class models don't usually provide new information over and above latent or multi-level growth. People just misunderstand them as doing so because they get more lines. Because they forget that growth models estimate intercepts and slopes that vary across ppts too. 1140
Andrea Howard @drandreahoward.bsky.social · 20/06/2025THE THING IS, the ABCD sample is large enough you could actually create cross-validation samples and test the robustness of the results within the same sample!! 130
Richard D. Morey @richarddmorey.bsky.social · 20/06/2025Came across one of these recently on the cross of levels of discrete time (3 points!) and discrete levels of the DV (5, maybe?). transparently bad, and of course just basically reproduced the patterns you mention. 020
John Sakaluk @johnsakaluk.bsky.social · 20/06/2025(BTW: never heard the "salsa pattern" terminology before--it's incredible--is there a go-to reference for that?) 140
