This effort builds on years of puzzling through informatics and practical data to build models. You've seen lots of amazing projects using these data types, and hopefully this facilitates more in the future! FAIR principles matter.

In this preprint, we describe this head-scratcher of a devs challenge, and present an information standard designed to be sufficient, not onerous, and flexible. We also illustrate (nicely done, @phuxley.bsky.social) the value of disaggregated data for parameter estimation.

Traits are measured in experiments, or quasi-experimental design, mostly. But they can be many measurements of many things! So any data standard must be flexible, and information can be iterated. As we discovered over several years, developing VecTraits database doi.org/10.7274/2802...

But the data are reported in so many different ways in the literature, sometimes in aggregate, sometimes with some units specified, and others not... this makes comparisons and synthesis (think replication and reuse), complicated.
What is the minimum amount of information needed?