brianstock
commented
6 years ago Hello,
Sorry for the slow reply! I haven't built in that functionality for TEF data. From a modelling standpoint, I think you could justify either estimating longer term diet separately or jointly with recent diet. In general, using all data in one model will give you tighter CIs, unless the data support different p estimates. And jointly is also probably more justifiable (because long term diet and recent diet are related).
Are you using raw source data or summary stats (mean/SD/n)? Because MixSIAR simply adjusts the source data by the TEF values, if you have summary stats, you could run a model with all data by:
- mix data: include "Tissue" as a column with values "Liver" and "Muscle". Specify "Tissue" as a fixed effect via
factors="Tissue", fac_random=FALSEinload_mix_data. - source data: calculate mean/SD/n for all sources, then add the "Liver" and "Muscle" TEF means/SDs to these, including "Tissue" as a column. Then specify
source_factors="Tissue"inload_source_data.
- TEF data: since you've adjusted the source data manually, include a TEF datafile with all 0s.
Separate models are easy to run, you just have to split your data into different files. However, this makes the assumption that the diet based on liver samples is independent of the diet based on muscle samples...
DEHewitt
Hi Brian! So I have juvenile Atlantic salmon from 2 sites, each site is a model and I have pooled the prey into FFGs. All works well ! But.. I've sampled both liver and muscle tissues from these fish. Assuming that I want to use the best TEF values out there, liver and muscle fractionnation differs. Can I have my discrimination file all incorporated in the 1 site model with two tissues with contrasting TEFs? Or, do I have to have 2 models per site (1 for recent diet ie. Liver) and another for longer term diet ie. muscle ?