concentration dependent models

[rivkachasan]

I am trying to form mixing models based on my GC-C-IRMS results of the c16.0 and c18.0. Following the geese example, to make a concentration dependent model I need to add 2 columns in the sources file titled Concc16.0 and Concc.18.0 but when I try to load the source information, I the receive the following error: Error in load_source_data(filename = source.animal3, source_factors = NULL, : Concentration dependence column names mislabeled. Should be 'Conc' + iso_names, e.g. 'Concd13C' if iso_names = 'd13C'. Please ensure Conc headings in source data file match iso_names in mix data file and try again. Alternatively, you may have set conc_dep=T by mistake.

Does anyone know how to work around this? I assume it is confused by the labelling

Thanks in advance for your help!

Rivka

[brianstock]

It would help if you post the first few lines of your mix and source files, as well as the code you're trying to run. Otherwise, you can look at lines 100-112 of load_source_data() to see exactly what it expects and in which cases it throws that error: https://github.com/brianstock/MixSIAR/blob/master/R/load_source_data.R#L100.

[adbreu]

Hi Rivka!

It looks like you are using the ∆¹³C values as a third isotopic value in a concentration dependent model. Therefore, the model expects the concentration of the ∆¹³C, but that doesn't exist! After all, ∆¹³C is just δ¹³C18:0 - δ¹³C16:0.

Most of the papers I've seen using Bayesian mixing models for pottery organic residues do not use the ∆¹³C values (Fernandes et al 2017, Lucquin et al 2018, Cubas et al 2020, Courel et al 2020). If you feel the ∆¹³C must be used, you can always make a not concentration depdendent model, but that will clearly affect the results.

Beyond this specific problem, it also looks like you're working with samples from the Middle East (which is really cool! can't wait for the publication!), I'm sure you are aware of the work of Gregg et al 2009, you might want to reconsider the mean and standard deviation of your sources, as reference fats from more arid and hot climates have been shown to present more enriched isotopic values. In fact, your "consumer" data has significantly high isotopic values (Tel Tsaf 18.24 -18‰ WOW!). You might want to consider other sources such as marine, C4 plants or C4/marine fed animals. Cynthiane Spiteri's PhD from York in 2012 reports C4 reference values, also check Dunne's work in 2012 from Kenya.

I hope this helped, all the best, Adrià Breu

[brianstock]

Sorry for the slow reply... Yes, there's an issue or two with your source file. As the error message says, your concentration dependence column names are mislabeled, should be 'Conc' + iso_names. You have c16.0 in iso_names from the mix file, so the corresponding column label should be Concc16.0 in the source file, not Conc.c16.0 like you have. It's easy to make a mistake like that, helpful to have another set of eyes look at it!

If that doesn't fix it, you might have an issue because your source file has concentration data for 1 of the 3 tracers. You might need to add columns Concc18.0 and Concd13C with all 1s. This will use concentration dependence for Concc16.0, the non-1 column you already have, and not for the other 2 tracers.

[ncttrinh]

Hi Brian,

I have a quick question and I am not sure it it is explained clearly elsewhere, including the manual. Can you explain briefly what "concentration dependent" mean? Does it mean the consumer (or the "sink", if it is a more appropriate word) isotopic value depends on the concentration of the sources? How do you determine if it is dependent or independent? In the Geese example, I can see in the "discrimination" excel file, there are mean +/- SD of the sources, so I assume the "concentration dependent" is related to bioaccumulation, am I correct?

Thanks hugely for your reply, Brian :)