SISUS papers

From Community

Here's a plan for the order of upcoming manuscripts for Stable Isotope Sourcing.

1. Have the first four (MS, LC, PSC, Bias) go out together, maybe to appear back-to-back in Oecologia.
2. Next, frequentist methods (Fvar, Freq)
3. Then, into Bayesian territory (Bayes)
4. More sophisticated models can follow (MN)

Contents 1 SISUS Models and Software (MS) 2 Linear Constraints (LC) 3 Proportion of Sources Consumed (PSC) 4 Bias Associated from not including CD and AE (Bias) 5 General Frequentist solution for variation in unique solution case (Fvar) 6 Likelihood of Ap=b=0 model (Freq) 7 Bayesian modeling (Bayes) 8 Macronutrient model (MN) 9 JASA Applications paper

[edit] SISUS Models and Software (MS)

(nearly complete)

• Describes synthesis of models (through AECDMM) in a common linear structure
• Describes sampling algorithm
• Introduces software to perform analysis

[edit] Linear Constraints (LC)

• short paper about prior belief and modeling using linear constraints
• also visuals of what some linear constraints do to the solution polytope

[edit] Proportion of Sources Consumed (PSC)

• With a good example dataset can show the extension to this model

[edit] Bias Associated from not including CD and AE (Bias)

• pull a few examples from the literature that use the BMM
• obtain associated information for CD and AE for the sources and mixture in the model
• Reveal the bias due to not incorporating CD and AE
• Perhaps easy with bear data
  • Grant Hildebran works with Charlie T. Robbins at the Bear Center. book: "Wildlife Nutritional Ecology".
  • Washington State U, nutritional ecologist studying bears.
  • Bear sources (ae): berries (sugars and lipids, 100%), salmon (all protein, high), grass (c3 low)

[edit] General Frequentist solution for variation in unique solution case (Fvar)

• General formula for variation when unique solution, delta method paper
• Ap=b has A and b subject to error
• What if know b exactly but A approximately?
• Explain what can and can not be done, and why.
• via Implicit Function Theorem?

[edit] Likelihood of Ap=b=0 model (Freq)

• Frequentist as far as it can go

[edit] Bayesian modeling (Bayes)

• On into the future

[edit] Macronutrient model (MN)

• Different tissues consumed as part of the biomass are preferentially routed to those tissues in the mixture.
• So each source has a number of tissues each with its isotope values, concentrations, and assim effic.
• Also, these are split up and assimilated into the tissues in the mixtures
• This level of realism is very complicated

[edit] JASA Applications paper

Show the stat stuff with the cool ecology application.