Jasmine Vieri
@jasminevieri.bsky.social
110 followers 90 following 9 posts
Computational archaeologist with a soft spot for all things metal. Postdoc @ University of Cambridge
Posts Media Videos Starter Packs
jasminevieri.bsky.social
El material suplementario incluye un resumen extendido en español: ars.els-cdn.com/content/imag...
ars.els-cdn.com
jasminevieri.bsky.social
All of the data and code are fully reproducible and publicly available on a dedicated repository: github.com/jmkvieri/BBLoP

Funded by AHRC, ERC (@reverseaction.bsky.social), and the Osk. Huttunen foundation.
jasminevieri.bsky.social
6️⃣The new modelling tools are readily applicable beyond archaeometallurgy, e.g, to compositional studies on ceramics or glass, or even to modelling the variability of diets in isotopic studies
jasminevieri.bsky.social
5️⃣The new tools were applied to the study of Muisca goldwork from pre-Hispanic Colombia (AD 600-1600). The findings suggest the intra-regional circulation of imported gold, with people pooling metals from a variety of geological sources for communal festivities.
jasminevieri.bsky.social
4️⃣Hierarchical (multilevel) model specifications can simultaneously consider both local and regional patterns in past craft production activities, whilst also accounting for sampling uncertainty. Think of it as zooming in and out on the past!
jasminevieri.bsky.social
3️⃣Variability in dispersion can tell us about behavioural factors in the past—like centralised control over resources or local improvisation. For example:

High dispersion ➡️ decentralised or improvised production
Low dispersion ➡️ more standardised practices
jasminevieri.bsky.social
2️⃣Compositional averages are often used to identify the desired performance characteristics of materials or in reconstructing artefact provenance.

But what if compositional dispersions are just as revealing?

We move beyond averages in explicitly defining 4 main sources of compositional variability
jasminevieri.bsky.social
1️⃣Compositional data are often skewed and heteroskedastic, making them challenging to model.

Traditional approaches within archaeology, such as simple linear regression on log-transformed data, can make numerically impossible predictions.

Beta regression is shown to be a more robust alternative