I have finished the parameter estimation report (sort of). I decided not to reconstruct the complex set of analyses that went into calculating the RIL-specific Gompertz parameters in the runup to the Science paper (which didn’t get used). This will probably have to be done if a version of this document goes into a manuscript submission, but I’m confident that the results won’t change and I should move on to running simulations and writing a paper.
Next steps
In the run-up to the Davis talk, I generated a bunch of simulation results for Ler, but crashed and burned on RIL. In the attempt to get the RIL model working, I may have compromised the Ler model (I was trying to have one model for everything, I think). So here’s what I need to do:
- Check that the model still runs as a single-genotype model
- Evaluate whether I need to rerun Ler (did parameter estimates change? Did I run enough replicates?)
- Evaluate whether I need to “validate” the Ler model against anything other than spread rate mean and variance
- Ensure that the codes for running the models are in reproducible scripts, and that the model functions themselves are adequately documented
- Get the multi-genotype models working
- Run RILs with and without evolution
- Run single-genotype model with mean RIL params to get a baseline for heterogeneity
I’ve checked a few things. For (1), there is still some debugging to do, as the run I just did eventually generated a -Inf value for runway_end.
For (2), yes, I need to re-run, as I’ve added a limit to the number of new pots, based on the description of the experiment.