Forgot the paper link! That's here doi.org/10.1111/ddi....

Lynn Dicks, Cang Hui, Iwan Jones, Ben Woodcock, @drnickisaac.bsky.social, and @andypurvisnhm.bsky.social

Re further data collation, for anyone that happens to be at EntSoc in Portland, look out for Eliza Grames' symposium on evidence synthesis.

And my brilliant co-authors on team GLiTRS, Grace Skinner, Andrew Bladon, @r-cooke.bsky.social, @charlieouthwaite.bsky.social, @james-rodger-za.bsky.social, Lindsey Barnes, Justin Isip, Daero Keum, Cristina Raw Villarino, Emily Wenban-Smith

Lastly, some thanks! Massive thanks to our funders, the NERC Highlights Grant ‪‪@glitrs.bsky.social, and more recently @leverhulme.ac.uk and The Isaac Newton Trust on my ECF.

I’m conscious that although this has been hard, performing experiments and collecting data is also hard. I’ve seen my role as to help lift up the work of others. For those that use these data, please do also cite the associated meta-analyses.

This database represents a huge collaborative effort. That’s true of the GLiTRS team, with a lot of this work having been performed by early career researchers, but it’s also true of the research we have built on.

Some of this logic you can read in our recent publication (lnkd.in/eRiXUcqf), led by @r-cooke.bsky.social and @charlieouthwaite.bsky.social

That means each effect size needs a structure such that it’s internally consistent with the meta-analysis within which it falls, internally consistent with another distinct meta-analysis, and externally consistent with predictions from two other evidence types.

But that’s not all, there are more moving parts. The intent has been to build such that we can combine this database with space-for-time and expert elicitation predictions.

To ensure all effect sizes are collected consistently, we we needed a protocol to guide two forms of effect size collation: first, the collation of effect sizes from previously published papers; and second, the collation of effect sizes from meta-analyses we performed ourselves (see osf.io/jw3gh)

Even the most robust data structures are only as strong as that structure is followed. That means Dynameta alone was not enough.

The intent on Dynameta was to build software that could ingest effect sizes among all possible combinations of IUCN threats and insect Orders. The novelty for me is not its interactivity, but the lego-brick back-bone it imposes on its user.

Dynameta is an interactive meta-analytic platform for performing ecological meta-analyses (lnkd.in/eu7QDBmX). That says what it does, but it doesn’t say why we needed it.

What I've tried to aim for here is the highest level of meta-analytic synthesis on insect biodiversity change. That intent meant starting this project off 3 years ago with two core structures: First, Dynameta; and second, a meta-protocol that guides among all meta-analyses.

How can you take X meta-analyses on insect biodiversity change, performed in many regions of the world in many taxonomic groups, and for all the ways in which humans moderate the environment, and then collate these estimations meaningfully among many researchers?

This is the toughest piece of work I've undertaken. Evidence synthesis is so hard to get right, even for an individual research question. But here the task was harder: