Alligator example

[Daniel-Gorman]

Hi Brian, fairly new user of MixSIAR. I am trying to run the script for the recently included Alligator example through R Studio, but I don’t seem to be able to get past line 173. This then appears to mean that I cannot subsequently ‘compare_models’ (screenshot below). Apologies if it’s a naively simple issue, but any advice would be greatly appreciated.

Attached also is Sessioninfo():

[brianstock]

Hi Daniel,

I just re-ran mixsiar_script_alligator.R on my machine to confirm it is working.

My guess is you're getting an error in the for loop with the call to plot_data (since it's graphics-related) and then not actually running the models. If you look at each element of the jags.mod list, you should see rjags objects. I.e. if you run class(jags.mod[[1]]), the output should be [1] "rjags".

Assuming this is the case, set mod=1 and then go through the for loop line-by-line to see what is causing the error. My guess is that it is related to graphics. Sometimes R Studio causes issues here, so first thing I'd try would be using the R console instead.

[Daniel-Gorman]

Hi Brian, Ive checked the objects - and get [1] = NULL (rather than "rjags").. ?
Also, I'm not quite sure what you mean by "go through the for loop line-by-line"..
Thanks for any additional help.

[brianstock]

Yeah, that's what I expected - you're getting an error before running the model (there is no model object saved as jags.mod[[1]]). After setting mod=1, run the lines inside the for loop (i.e. 109-166) one command at a time so you can see where the error occurs.

[Daniel-Gorman]

Hi Brian, slowly getting through the examples. Managed to fix the previous issue by changing the for loop section:

plot_data(filename=paste0("isospace_plot_", mod),
plot_save_pdf=FALSE,

I'm now using my own data (with 4 competing models), but getting the following error when moving back up to the root directory:

[brianstock]

Good to hear! Saving PDFs is likely an OS-specific issue. Did you try using the R console instead of R Studio?

The new error is happening reading in the source CSV file, not when moving back to the root directory. It's probably worth a bit of your time to do an R tutorial and understand what a for loop is. Cyclismo is pretty good: https://www.cyclismo.org/tutorial/R/scripting.html#for-statements.

[Daniel-Gorman]

Hi again Brian, making steady progress now :) Thanks for the very responsive help..
Ive made it all the way through running my 4 models (Null, factor, covariate, factor+covariate) and used LOO to compare the weights!
Now trying to plot the covariate (similar to the Alligator plot) but bumped into some more issues. Any advice on the following error?

[brianstock]

Sorry, this error is sloppy coding/testing on my part. The calc_eps function should be moved above where it is called. I've fixed this on the develop branch, which you can install via

devtools::install_github("brianstock/MixSIAR", ref="develop")

Alternatively, you can just copy+paste the calc_eps function (lines 274-280), then re-run that code.

[Daniel-Gorman]

Hi Brian - this worked - thanks greatly.
So, now I’m almost where I want to be, but I fear that the "hard part" is yet to come. My covariate (‘distance’) model includes 6 dietary sources (which makes interpretation very difficult). What I would like to do is combine the 6 sources – into 3 groups (a posteriori)..
Something like this (taken from the Mantis e.g.):

combined <- combine_sources(jags.model, mix, source, alpha.spec,
groups=list(group1=c("source1","source2"), group2=c("source3","source4"), group3=c("source5","source6")))

..and then Plot [group] proportions vs. ‘distance’ from covariate model.
I'm currently trying to merge the code from Mantis (combine) + Alligator (plot proportions vs. length) > but not getting very far :0
Any advice on how to move forward, greatly appreciated.