I want to propose the addition of a function to find endmembers using PCA analysis and geochron slope alignment, as published in Shnyr et al (2026).
Shnyr, E., Kuflik, T., Desai, K., & Eshel, T. (2026). Determining the origins of Phoenician silver: Exploring the potential of machine learning for lead isotope analysis. Journal of Archaeological Science, 188, 106499. https://doi.org/10.1016/j.jas.2026.106499
Description
We need to implement a new method for endmember expansion and classification within our lead isotope analysis suite. While standard PCA (via statprcomp) identifies the extreme points ($p_a$ and $p_b$) along the first principal component ($c_1$), these two vectors often fail to capture the full geological heterogeneity of ore sources.
Following Albarède et al. (2024b) propose a method to expand these endmembers by identifying points that align parallel to the Geochron.
Proposed Workflow
- Initial PCA: Perform PCA on 3D isotope data ($^{206}Pb/^{204}Pb, ^{207}Pb/^{204}Pb, ^{208}Pb/^{204}Pb$).
- Primary Identification: Define $p_a$ at $\min(c_1)$ and $p_b$ at $\max(c_1)$, provided $c_1$ accounts for $>95%$ of total variance.
- Geochron Expansion: Identify additional endmembers $e_i$ that satisfy the proximity to the Geochron slope ($S \approx 0.626208$).
- Condition: A vector $\vec{e}_i$ is an endmember if it falls within tolerance $\epsilon$ of the line $p_i$ parallel to the geochron in the $(^{206}Pb/^{204}Pb, ^{207}Pb/^{204}Pb)$ plane.
- Equation:$$|e_2 - (p_1 + (S \times (e_1 - p_1)))| < \epsilon$$Confidence Threshold: Define sets $E_a$ and $E_b$. A set is considered "confirmed" only if $|E_i| > 2$. If $|E_i| \le 2$, the point is flagged as a "possible endmember" for manual review.
Requirements
Function Name: expand_geochron_endmembers()
Parameters:
-
data: 3D lead isotope data frame.
-
epsilon: Tolerance value (default 0.01 based on empirical results).
-
slope: Geochron slope (default 0.626208).
Outputs: A classified list of confirmed vs. possible endmembers.
Calculated probabilities for provenance based on isotopic similarity to existing ore databases.
I want to propose the addition of a function to find endmembers using PCA analysis and geochron slope alignment, as published in Shnyr et al (2026).
Shnyr, E., Kuflik, T., Desai, K., & Eshel, T. (2026). Determining the origins of Phoenician silver: Exploring the potential of machine learning for lead isotope analysis. Journal of Archaeological Science, 188, 106499. https://doi.org/10.1016/j.jas.2026.106499
Description
We need to implement a new method for endmember expansion and classification within our lead isotope analysis suite. While standard PCA (via statprcomp) identifies the extreme points ($p_a$ and $p_b$ ) along the first principal component ($c_1$ ), these two vectors often fail to capture the full geological heterogeneity of ore sources.
Following Albarède et al. (2024b) propose a method to expand these endmembers by identifying points that align parallel to the Geochron.
Proposed Workflow
Requirements
Function Name: expand_geochron_endmembers()
Parameters:
data: 3D lead isotope data frame.epsilon: Tolerance value (default 0.01 based on empirical results).slope: Geochron slope (default 0.626208).Outputs: A classified list of confirmed vs. possible endmembers.
Calculated probabilities for provenance based on isotopic similarity to existing ore databases.